KR102397791B1 - Apparatus and method for automatically converting note to action reminders - Google Patents

Abstract

The present disclosure provides a method for automatically converting a note-to-self sent from an electronic device to an action reminder. The method includes the steps of: detecting an input including at least one vocabulary from a user of the electronic device; It includes the process of creating a reminder based on the context of the analyzed surrogate expression or referential expression.

Description

APPARATUS AND METHOD FOR AUTOMATICALLY CONVERTING NOTE TO ACTION REMINDERS

In general, the descriptions below relate to automatically generating a reminder, and more specifically, to automatically convert a note-to-self (N2S) sent from an electronic device to an action reminder. It relates to an apparatus and method.

The automatic reminder generation method provides a reminder based on text input as a note. The automatic reminder generation method mainly extracts text information from a template or keyboard-based user input. Accordingly, information extracted for creating a reminder is limited to the area of the template and keyboard.

The following descriptions are intended to provide an apparatus and method for automatically converting a note-to-self (N2S) sent to the electronic device into an action reminder based on a semantic representation and context.

An embodiment provides a method for automatically converting a note-to-self (N2S) sent from an electronic device to an action reminder. The method includes the steps of: detecting an input including at least one vocabulary from a user of the electronic device; analyzing an anaphora expression or a dictic expression for each of the at least one vocabulary; It includes the process of creating a reminder based on the context for the substituted or referential expression.

An embodiment provides an electronic device for automatically converting a note-to-self sent to the user into an action reminder. the electronic device is configured to detect an input comprising at least one vocabulary from a user of the electronic device, and is configured to analyze an anaphora expression or a dictic expression for each of the at least one vocabulary; It is configured to generate a reminder based on the context of the analyzed surrogate or referential expression.

One embodiment provides a computer program including computer executable program code recorded in a readable non-transitory storage medium. The computer-executed program code includes, by an input recognition unit, detecting an input comprising at least one vocabulary from a user of the electronic device, and an anaphora for each of the at least one sentence. Alternatively, it is recorded on the medium in order to execute the steps of analyzing the directive expression and generating a reminder based on the context for the analyzed surrogate expression or referential expression.

The above embodiments and other embodiments can be understood in conjunction with the detailed description and drawings described below. However, the detailed description set forth below is illustrative means for describing preferred embodiments and various details thereof, and does not limit the scope thereof. Various modifications and variations may occur within the scope of the above embodiments.

The following descriptions may provide an apparatus and method for automatically converting a memo sent to the user from an electronic device into an action reminder.

For a more complete understanding, the following descriptions are made with reference to the accompanying drawings. In the drawings, like reference numbers indicate like elements.
1 illustrates a functional configuration of an electronic device for automatically converting a note-to-self (N2S) sent to the user into an action reminder.
2 shows a flow of operations for generating a reminder.
3 shows a flow of operations for determining a context based on a plurality of semantic representations.
4 illustrates a flow of an operation for merging at least two contexts to determine one context.
5 illustrates a flow of an operation for preprocessing a memo sent to the user.
6 shows a flow of functional operations of a Language Engine (LE) unit.
7A shows an example of an operation for converting a user input into a canonical representation.
7B shows an example of a syntactic parsing operation.
7C shows an example of an operation for generating a semantic representation.
7D illustrates an example of an operation for extracting a context.
7E shows another example of a syntax parsing operation.
7F shows another example of an operation for generating a semantic representation.
7G shows an example of the operation of merging semantic representations.
7H shows another example of an operation for extracting a context.
7I shows another example of a syntax parsing operation.
7J shows another example of an operation for generating a semantic representation.
7K shows another example of the operation of merging semantic representations.
7L shows another example of an operation for extracting a context.
7M illustrates an example of an operation of automatically converting a memo sent to me into an action reminder based on a semantic expression.
8 shows an example of an operation of automatically converting a memo sent to me into an action reminder.
9 shows another example of an operation for automatically converting a memo sent to me into an action reminder.
10 illustrates a computing environment implementing a method for automatically converting a memo sent to me by an electronic device into an action reminder.

Terms used in the present disclosure are used only to describe specific embodiments, and may not be intended to limit the scope of other embodiments. The singular expression may include the plural expression unless the context clearly dictates otherwise. Terms used herein, including technical or scientific terms, may have the same meanings as commonly understood by one of ordinary skill in the art described in the present disclosure. Among the terms used in the present disclosure, terms defined in a general dictionary may be interpreted with the same or similar meaning as the meaning in the context of the related art, and unless explicitly defined in the present disclosure, ideal or excessively formal meanings is not interpreted as In some cases, even terms defined in the present disclosure cannot be construed to exclude embodiments of the present disclosure.

In various embodiments of the present disclosure described below, a hardware access method will be described as an example. However, since various embodiments of the present disclosure include technology using both hardware and software, various embodiments of the present disclosure do not exclude a software-based approach.

Hereinafter, the present disclosure will describe a device method for automatically converting a memo sent to the user by an electronic device into an action reminder. Prior to the description of the present disclosure, definitions of terms are described as follows.

In the present disclosure, the canonical representation may mean a modified expression of a natural language (eg, a language used in chatting) to a predetermined format (eg, an official Korean dictionary). In the present disclosure, a semantic representation may mean information that a specific vocabulary or sentence has. In the present disclosure, a context is a linguistic expression revealed in light of an internal context (ie, a situation revealed in light of the written sentence and the preceding and following sentences) or external context (ie, a situation known other than the written sentence) of a given sentence or vocabulary. It can be context or meaning.

The terms defined above and terms used below are exemplified for convenience of description. Accordingly, the present disclosure is not limited to the terms described below, and other terms having equivalent technical meanings may be used. For example, the reminder may be referred to as a notification, an alarm, schedule information, or the like.

1 illustrates a functional configuration of an electronic device 100 for automatically converting a note-to-self sent to the user into an action reminder.

1 , an electronic device 100 includes an input recognition unit 102, a preprocessing unit 104, a language engine (LE) unit 106, a context engine (CE) unit 108, and generalized ontology 110. The input recognition unit 102 is configured to detect an input from a user of the electronic device 100 . In an embodiment, the input may be in the form of text. In another embodiment, the input may not be in text form. For example, the text input may be text information such as alphabets, words, and numbers. As another example, the non-text input may be a voice input. Automatic speech recognition (ASR) may be one of the input recognition methods for converting the speech input into text information. Other input recognition methods may be optical character recognition (OCR) and hand written character recognition (HCR) methods.

The pre-processing unit 104 is configured to transform the input into a canonical presentation. As described above, the normative representation may mean an expression obtained by modifying a natural language according to a predetermined format. For example, the pre-processing unit 104 may correct slang words and typos into words written in an official language dictionary. The pre-processing unit 104 disambiguates a plurality of corrective words and differences for the slang and typo based on the local context. The pre-processing unit 104 tokenizes one or more word segments representing a unit of a single entity, an activity, a spatial representation, a temporal representation, and the like. In the present disclosure, tokenization may refer to a process of discriminating each of the plurality of words in a sentence including the plurality of words. Further, the pre-processing unit 104 generates a normative representation of the temporal representation.

The language engine (LE) unit 106 is configured to extract information from the normative representation to generate a plurality of semantic representations. As described above, the semantic expression may mean information of a specific vocabulary or sentence. The LE unit 106 obtains semi-tokenized sentences from the pre-processing unit 104 as input. Also, the LE unit 106 extracts relevant information from the semi-tokenized sentences. For example, the semi-tokenized sentences may be activity type, purpose, participant, time, location, etc. The LE unit 106 generates semantic expressions for the input sentences, and transmits the generated semantic expressions to a context engine (CE) unit 108 . The process of generating the semantic expression depends on the language structure. However, the semantic expression may be independent of the language structure. The semantic expression depends on the semantics of the word and how related the words are to each other. Examples having the same semantic expression are as follows.

1. CC cost up to 4 days. (CC Bill is due on 4th.).

2. You must pay the CC fee by the 4th day. (Need to pay CC bill by 4th.).

3. 4 days is the last day to work out CC costs. (4th is the last date to settle CC bill.)

4. Don't forget to pay the CC fee by the 4th day. (Don't forget to pay for the CC bill by 4th.)

5. The last day to settle the CC cost is 4 days. (Last day to settle CC bill is 4th.)

The LE unit 106 may extract relevant information from any natural sentence structure type (eg, a sentence structure used in chatting). The main features of the LE unit 106 are as follows.

1. Extract the main activity type and related activities from the input sentence.

2. Extract the participant list from the input sentence based on the rules related to the activity. In an embodiment, the participant may be an agent, a patient, and a co-actor. For example, in a sentence written "John will wish Tom on his anniversary.", John may be identified as the actor of the activity, and Tom as the subject of the activity.

3. Extract the purpose of the activity from the sentence.

4. Extract the time and location of the activity from the sentence.

5. Distinguish between unrelated or separate times and locations in the same sentence. For example, suppose you have a sentence that says "meet John on Monday at 4pm and Tuesday at 6pm". The proposed method can distinguish the sentence as "meet John on Monday at 4pm and meet John on Tuesday at 6pm". Here, based on the identified different times, reminders may be set for 4 pm and 6 pm, respectively.

6. Infer and extend a common temporal or spatial representation through two unrelated instances in the same sentence. For example, in a sentence written "let's meet John on Mondays and Tuesdays at 6pm", 6pm is inferred as the common time for all of the above meeting dates, and each object propagated to (object).

7. Infer the semantic expression of the sentence, and transform the activity type according to the criterion. For example, in a sentence written "I will see John on Monday", the activity type would be converted from 'see' to 'meet'.

8. Extract intent from sentences that include questions, suggestions, information transfers, etc.

9. Extract tense information from sentences.

The CE unit 108 is configured to determine a context based on the plurality of semantic representations. The context may be a linguistic context or meaning revealed in light of an internal context (i.e., a situation revealed in light of the written sentence and preceding and following sentences) or an external context (i.e., a situation that is known other than the written sentence) of a given sentence or vocabulary. there is.

The CE unit 108 captures and stores the sentences in the memo. In addition, the CE unit 108 analyzes anaphora and dictic expressions in a sentence and performs temporal and spatial reasoning among other functions. Here, the substitute expression may mean an expression that can be used in place of a word already in the sentence, and the referential expression may mean an expression referring to a specific object. For example, if there is a sentence that says "John is going to London. I have to meet him there next week", in that sentence 'him' can be a substitute for 'John' used in the previous sentence. As another example, if there is a sentence written "I like this flower.", 'this' in the sentence may be a referential expression referring to a flower.

The generalized ontology 110 is a knowledge base constituting diversified information such as a slang dictionary related to a specific area, an occurrence dictionary related to a time, and entity and rule names. The generalized ontology 110 may also store information such as vocabulary (words and parts of speech (POS) tags, frequency of occurrence, etc.) and word mapping (mapping words into semantic concepts). The generalized ontology 110 can be accessed by modules for different information and data. In addition, the generalized ontology 110 may capture the knowledge of the specific area by describing concepts in the specific area that are the main interest and the relationship between the concepts. For example, the slang dictionary may include items that can be used by the pre-processing unit 104 (eg, tomorrow is expressed as Tmrw). Also, spatial information including spatial relationships and rules for gender assignment and the like may be used by the CE unit 108 .

1 illustrates limited embodiments of the electronic device 100, other embodiments may be applied. The labels or names of the units are for illustrative purposes only and do not limit the scope of the present invention. Also, the electronic device 100 may include any number of units or sub-units. The units or sub-units may communicate with each other along with other components. Also, one or more units may be combined into one unit or divided in other ways without departing from the scope of the present invention.

2 shows a flow of operation 200 for generating a reminder.

Referring to FIG. 2 , in step 202 , the electronic device 100 detects an input including at least one vocabulary from a user of the electronic device 100 . A method of detecting the input may vary according to an implementation method. For example, the electronic device 100 may detect the input using OCR, HCR, etc. in addition to voice recognition.

In step 204, the electronic device 100 parses at least one of a surrogate expression and a referential expression for each of the at least one vocabulary. A detailed method of parsing at least one of the substituted expression and the referential expression will be described later in detail with reference to FIG. 4 .

In step 206, the electronic device 100 generates a reminder based on at least one of the parsed substitute expression and referential expression. The user of the electronic device 100 may generate the reminder by using at least one of the parsed substitute expression and referential expression to create a reminder in which the user's intention is more clearly revealed.

3 shows a flow of operation 300 for determining a context based on multiple semantic representations.

Referring to FIG. 3 , in step 302 , operation 300 includes detecting an input from a user of the electronic device 100 . In other words, in step 302, the electronic device 100 detects an input from the user. In an embodiment, the input may be in the form of text. In another embodiment, the input may be in a form other than text. For example, the input may be text information or a voice input.

In step 304, the operation 300 includes converting the input into a normative representation. In other words, in step 304 , the electronic device 100 converts the input into the normative representation. For example, the pre-processing unit 104 may convert 'Wed' into 'Wednesday'.

At step 306, the operation 300 includes extracting information from each of the normative representations of the input to generate a plurality of semantic representations. In other words, in step 306, the electronic device 100 extracts information from each of the normative representations of the input to generate a plurality of semantic representations. The generation of the semantic expression is based on the meaning of the syntax and word vocabulary of the language in the sentence and the relationship between them. For example, the relevant information may include activity type, intent, participant, purpose, time, location, and the like.

In step 308, the operation 300 includes determining a context based on the plurality of semantic expressions. In other words, in step 308 , the electronic device 100 determines a context based on the plurality of semantic expressions. For example, there is a sentence that says "John is going to London. I have to meet him there next week". In this case, the meeting location in the sentence will be interpreted as "London", and the participant will be interpreted as "user (N2S or user of the electronic device 100)" and "John". In the above sentence, the meeting time starts with the first day of the week returning based on the reference time, and sets the last day of the week as the end, and each interval is an open interval. will be set The reference time may mean a user's input time.

4 illustrates a flow of operation 400 for merging at least two contexts to determine one context.

Referring to FIG. 4 , in operation 402 , operation 400 includes a process of resolving a surrogate expression and a referential expression in the plurality of semantic expressions. In other words, in step 402, the electronic device 100 identifies a surrogate expression and a referential expression within the plurality of semantic expressions. The CE unit 108 analyzes the surrogate expression and the referential expression according to the following steps.

Indicative expression analysis in purpose: the CE unit 108 identifies a referential term. In addition, the generalized ontology 110 is requested to output a list of contexts including information on purpose. In addition, the most recent contexts are extracted, and contents related to the purpose of the extracted contexts are allocated to contents related to the purpose of the current context.

Parsing the referential expression in spatial: The CE unit 108 identifies the presence of the referential expression and the level of the referential expression. For example, when the sentence "I will meet you in that City" is written, the referential expression is related to the spatial level of the 'city'. The generalized ontology 110 is requested to output a context including non-empty spatial content. In addition, the most recently matched context having the same spatial level is extracted, and the extracted context is allocated to content related to the space of the current context.

Analysis of the referential expression of an occasion: The time can be analyzed into various types of occasions. For example, there are times related to food and times of annual festivals. In the case of food-related timing, the temporal information associated with all content-based "dates" is extracted from the matched context and populated with the current context for referential expression. For other temporal types, "year" information is extracted from the matched context. For example, if "Ram is coming to Bangalore in 2016. Shyam will meet him that Christmas." is written "that Christmas )" means Christmas in 2016. Another example: “Ram was in Bangalore on 16th January. We had lunch together. That lunch was excellent. ", "that lunch" means the lunch on January 16th.

Analysis of other temporal representations: In the case of other temporal representations, an appropriate temporal value is obtained from the matched context, and the acquired temporal value is filled in the temporal-related content of the input context. For example, if it is written "Ram is coming to town in Jan 2015. He will meet me that 16th." "that 16th" means January 16, 2105.

Analysis of substitute expressions and referential expressions for participants: In the content to be described later, a scoring-based model is implemented. The various parts of the algorithm are as follows. There is no separate chronology for the algorithm.

Analysis of first-person and second-person pronouns: The pronouns I, me, myself, you, yourself are independent of the previous context and are analyzed immediately at the beginning of the user information analysis. For example, in the previously mentioned sentence ("Tom is coming to Bangalore tomorrow. We will meet up on 15th"), "we )" may refer to users and some previously mentioned third party participants.

Gender matching: The first important criterion in the surrogate expression and referential expression for the participant is to match the gender using the generalized ontology 110. Various entities can be applied to the gender using the generalized ontology 110 . In this case, the ontology does not have an entity, and the LE unit 106 allocates the entity to a participant.

Cardinality Matching: Another important criterion in surrogate and referential expressions for the participant is that number should be considered while analyzing the cardinality match. In the present disclosure, the cardinality may mean the number of elements belonging to a specific set. References made in the 'plural sense' should be mapped to multiple entities, and references made in the 'singular sense' should be mapped to a single entity.

Distance Depreciation: As the contexts continue back one step at a time, the score for the factor decreases. If all other factors are equal, the last participant mentioned in the context will score the highest in number (n).

Location inference: The inferences are made by the CE unit 108 based on the travel information mentioned by the user. The inference is performed while maintaining details of travel information such as origin, destination, travel departure date, and end date. The above information applies to the various individuals referred to by the user. For example, "Mary will call me. Susan is going to Pune from the 12th to the 18th. I will meet her at the bungalow on the 15th. (Mary will call me. Susan is going to Pune) from 12th to 18th. I will meet her in Bangalore on 15th.)", "her" could be interpreted as both Mary and Susan. The system will interpret "she" as "Mary" because "Susan" is not in the bungalow on the 15th. The knowledge is obtained from travel information based on COMMON SENSE REASONING.

Gender table maintenance: When the electronic device 100 encounters unknown entities in the generalized ontology 110, it is necessary to specifically analyze a surrogate expression or a referential expression. This information is maintained by being stored in the same state in the "gender table". For each user, one table is maintained to ensure consistency. For example, if "Tiberius is coming to town. He will meet me." is written, information about "Tiberius" in the ontology is Since there are no other candidates for the referential pronoun "he", "he" will be interpreted as "Tiberius." Also, information that Tiberius is "male" is not found individually in the "gender table". Thereafter, "Tiberius" will not be interpreted as a woman.

In step 404, the operation 400 determines a degree of matching between at least two contexts based on at least one parameter. In other words, in step 404, the electronic device 100 determines the degree of correspondence between at least two contexts based on compatibility through activity, participant, time, location, purpose, etc. or a combination thereof. In one embodiment, the parameters may include activity, participant, time, location, purpose, etc. or a combination thereof.

The semantic expression generated for each sentence by the LE unit 106 may be analyzed by the CE unit 108 and merged with the context of the same memo. Searching the list of matched contexts identifies a suitable match based on different dimensions. A score representing the “degree of agreement” between the two variables is generated for each dimension based on compatibility. In this case, if some other intention or rejection is implied in the input, a high score will be output even if compatibility is low. Based on the score output by a sub-module (not shown) of the CE unit 108, a comprehensive score is generated. Based on the threshold value, the context with the highest score is identified as a matched context. Examples of the various dimensions are as follows.

Temporal matcher: The electronic device 100 matches time-related contents described in all matched contexts with time-related contents written in an input sentence. The module determines the score based on the degree of agreement between the two temporal contents. For example, "I am going to meet John on Sunday. I will meet him at Starbucks at 10 AM." In this case, 'Sunday and 10:00 AM' are indirectly matched.

Spatial matcher: The electronic device 100 determines a score based on a degree of matching between contents related to two spaces. For example, the former in "I am going to Whitefield on Sunday." and "Meeting with John at Phoenix Mall then" Device 100 determines "Phoenix Mall" as the location in "Whitefield". Thus, 'Whitefield' and 'Phoenix Mall' match.

Participant matcher: The electronic device 100 performs participant matching through matched contexts. For example, if it says "Meeting with John at the Forum on Sunday." and "He will reach Forum at 9 AM." , "he" can be interpreted as "John". The above sentences correspond indirectly to the participant.

Purpose matcher: The electronic device 100 determines a score based on a degree of matching between content related to two purposes. For example, if it is written "Discussion with John on Friday" and "I will discuss the X Company's proposal then." Company X's proposal' and the theme of the meeting indirectly coincide.

Occasion matching: The electronic device 100 matches content related to time in the input sentence with matched contexts. For example, "John's Marriage is on June 13th 2015." and "Attend his wedding" are "John's Marriage is on June 13th 2015." It's March 13. John's Marriage is on June 13th 2015. Attend his wedding." Match according to the compatibility of time between "marriage" and "wedding" in the above sentences.

Matching 'DoWhat': The 'what' refers to an object of a main action. In order to match the 'what', the electronic device 100 matches the content related to the 'what' described in the input context with the content related to the 'what' described in another matched context. For example, if "I have applied for Passport." and "I will receive it tomorrow." are written, the "applied" activity The 'what' for the activity is "passport", and the 'what' for the activity "receive" is "it". Here "it" is interpreted as "passport". Therefore, the contents of 'what' in the above sentence indirectly coincide.

Gap filling: The gap filling is used to fill in the missing information in the sentence based on the context of the memo (time stamp of the memo, user information, etc.) and the semantic representation of other sentences in the memo.

Default participant gap filling: The default participant gap filling is performed according to the following example. If "John will be in Total Mall. Remind to meet him." is written, "I/ME" in the semantic expression of the second sentence. is added as a participant.

Spatial gap filling: If the sentence contains an activity requiring a specific place, but there is no content for the specific place, the spatial gap may be determined by searching previous contexts. For example, if it is written "I am going to Pune. I will meet Tom.", as a place in the semantic expression of the second sentence above, "Pune )" is added.

Temporal gap filling: For example, written "John's birthday is on 25th. Remind me to call him." case, "25 days" is added as time information in the semantic expression of the second character.

In step 406 , the operation 400 includes determining one context by merging at least two contexts of the plurality of semantic expressions. In other words, in step 406 , the electronic device 100 determines one context by merging at least two contexts of the plurality of semantic expressions. The matched contexts are merged and updated in the context storage structure. Four primitive types of attribute merging operations are described below.

Append: The current attribute value is appended to the matched context attribute value. The contents of the attachment operation are as follows. For example, if the matched context is "I will meet Mary.", and the input context is "will meet Tom also.", in the context (I), Tom, and Mary are analyzed as participants by the attachment action.

Override: The current attribute value is overlaid on the matched context attribute value. For example, the matched context is "I will meet you at 5pm.", and the input context is "No, make it 6pm. (no, make it 6pm.) ", in the above context, 6 pm is analyzed as a temporal meaning by the overlapping operation (5 pm is overlapped by 6 pm).

Negate: The current attribute value is deleted from the matched context attribute value. For example, if the matched context is "I will meet Mary and Tom.", and the input context is "Tom will not come." , "I" and "Mary" in the above context are analyzed as participants by the invalidation action, and Tom is deleted from the final list.

No Change: The current attribute value is not merged with the matching context attribute value. For example, if the matched context is "Can we meet on 5th?" and the input context is "I will let you know tomorrow." , time information included in the input context should not be merged into the matched context. Each attribute is handled by a specific updater as follows.

Temporal updater: The temporal updater interfaces with a temporal speculator that processes temporal objects created by the LE unit 106 . The append action causes the two input periods in the time object to be merged and processed. In the overlapping operation, the structure and level in the section object are checked to generate a merged section object by merging two section objects. The interval objects are the result of processing to obtain the concept of time. For an invalidation operation, the CE unit 108 merely maintains a signal that a specific content (eg, activity, time, participant, place, purpose, etc.) in the semantic expression has been invalidated, which will later be superimposed with a higher probability. indicates that it can

AM/PM Analysis Criteria: Time Renewer analyzes AM/PM values based on activity type and time results. For example, in "Meeting on 5th with Mary at office at 2", the CE unit 108 does not specifically mention information about the meeting, although the user does not specifically mention information about the meeting. If not, infer that the encounter is more likely to occur at 2 PM than at 2 AM. Therefore, the time result is updated to 2:00 PM on the 5th. The AM/PM analysis is analyzed in the absence of evidence for a meridian of hour-minute time.

Spatial Updater: Spatial object merging is merged based on levels such as country, state, city, engine, and sub-engine. Level information of a known place can be accessed from the generalized ontology 110 . Also, ranges such as national, regional, and international are inferred from the level existing in the last spatial object.

Participant Updater: Each operation is performed on various participant lists, actors lists, actors lists, and other lists. In the append operation, if the participants are not on the list, the participants are added to the list. Invalidation removes the participants from the list. Overlay retrieves information from the context of the sentence, deletes the last added participants, and adds the participants created by the input.

The temporal deducer (a standalone component) handles complex temporal structures and provides a List of Result objects (ROs). The list of result objects includes the following list.

1. Date

2. Time

3. Time Interval

4. Whether the time interval is an open interval or a closed interval (Open vs Closed Interval of the Duration)

5. Repeat multi-level results. That is, the result is repeatedly generated based on the information extracted from the input structure. The above operation confirms that the date and time compatibility is inconsistent due to an erroneous mistake (Goof Up flag) in the input sentence.

For example, suppose you meet every Friday from February 1 to 28, 2016 for 8 hours. In this case, the time result object RO is as follows.

RO level-1

Start Time: 2016-02-01 00:00

End time: 2016-02-28 23:59

Repeat hops: P1W

Repeat Type: AND

RO level-2

Start Time: 2016-02-06 00:00

End time: 2016-02-06 23:59

Segment: PT8H

In the case of the above example, the input process produces a level-2 iteration result. The RO level-1 represents the outside of a boundary/interval in which repetition may occur. The hop means skipping on a weekly basis (P1W stands for 'every one week'). In other words, the interior of RO level-2 occurs within the bounds of the range on a weekly time basis. The repetition occurring every Friday may be indicated by setting one example (the first date) among Fridays of the outer boundary section. The types of every meeting and the occasional meeting can be distinguished by AND/OR/XOR. "every time" can be expressed as AND. The meeting time interval may be updated at RO level-2, and the interval is set to 8 hours in the RO level-2 (PT8H means 'every 8 hours'). The iteration occurs every Friday.

5 illustrates a flow of operation 500 for preprocessing a memo sent to the user. The operation 500 describes a process of pre-processing the text using the pre-processing unit 104 to extract time information. The N2S text preprocessing process includes the following process.

Referring to FIG. 5 , in step 502 , operation 500 includes correcting slang words and spellings into words in an official language dictionary. That is, in step 502, the electronic device 100 corrects slang words and spellings into words of an official language dictionary. For example, "2mrw", "tmrw", "tomorrowwwwwwww", "tomorrow", etc. are all corrected to "tomorrow".

In step 504, the operation 500 includes a process of clearly discriminating a difference between slang and spelling correction. That is, in step 504, the electronic device 100 distinguishes slang and spelling corrections based on regional contexts of the preceding word and the subsequent word. For example, "on January 4 4 4 hours" may be corrected to "on January 4 for 4 hours". Also, “in January 4 dinner” may be corrected to “in January for dinner”. Also, “between March 2 2 April 2” may be corrected to “between March 2 to April 2”. “Morning 2 noon” may be corrected to “morning to noon”.

In step 506, the operation 500 includes tokenizing one or more word fragments representing one unit. That is, in step 506, the electronic device 100 tokenizes one or more word fragments representing one unit. For example, in a sentence written "I will meet John in United States of America.", "United States of America" is a combination of a number of consecutive words, but A sentence represents a single unit. In one embodiment, the tokenized one or more word fragments are based on identification of temporal representations. For example, the pre-processing unit 104 identifies temporal representations as follows.

1. Calendar date (different from place indicating your location)

2. Hour-Minute Time (12 and 24-hour format)

3. Relative time (e.g. tomorrow, now, etc.)

4. An idiom related to the expression of time (eg, "see you in 5" means "see you in 5 minutes)."

5. Language is dependent on structure for temporal representation. (For example, "quarter past 3" means "3:15" and "5 to 7" means "6:55". The meaning of individual words may differ from the meaning of words used together as a whole.

6. Part of the day (eg morning, noon, evening, etc.)

7. Name of day of week

8. Name of month

9. Names of seasons

10. Names of national holidays and festivals

11. The name of the time period associated with the food (eg lunch, brunch, coffee, etc.). The time name associated with the food may or may not be associated with an existing time associated with it.

12. Name of period (eg birthday, marriage, etc.)

13. A name specifying the length of a time interval (eg, millennium, century, decade, year, quarter cycle, month, fortnight, week, day) , hours, minutes, seconds, etc.)

14. Repetition associated with each of the aforementioned temporal representations

15. Modifiers (e.g., ordinal and radix numbers, quantifiers (a, an, the, some, etc.) associated with each of the above temporal expressions), referential expressions (this, that, these, those, etc.), relative terms (next, last, subsequent, alternate, etc.), fragmentary modifiers (early, middle, etc.) mid), late, etc.), and other modifiers (more, less, half, double, etc.) is not)

In an embodiment, the pre-processing unit 104 may convert the identified temporal representation into a canonical representation. For example, the sentence "see u in 5" can be translated into "see you in 5 minutes".

6 shows the flow of functional operation 600 of the LE unit 106 .

Referring to FIG. 6 , in step 602 , operation 600 includes generating a parse tree. That is, in step 602, the LE unit 106 generates a parse tree. For each normative representation generated by the preprocessing unit 102, multiple parses are used to generate a syntactic representation (ie, a parse tree). Also, each parse tree is used to create a semantic representation.

In step 604, the operation 600 includes evaluating the accuracy of the parsing. That is, in step 604, the LE unit 106 evaluates the accuracy of the parsing. The evaluation of the parse tree includes the following steps.

1. A grammar-based rule for evaluating the accuracy of the pars is applied, and a score is assigned to each pars.

2. All n-grams tokenized by the preprocessing unit are not tokenized. The n-grams mean a probabilistic language model.

3. Based on different syntactic categories of respective words in the sentence, multiple semantic representations of each word are obtained.

In step 606, the operation 600 includes generating semantic representations. The LE unit 106 uses a plurality of semantic expressions for the word and generates a combination of the semantic expressions based on the score. Also, for all semantic expressions for the word, semantic objects are created.

In step 608, the operation 600 includes evaluating and scoring the semantic expressions. The LE unit 106 creates independently scored semantic representations for each combination by concatenating and scoring semantic objects. The LE unit 106 may connect the semantic objects using an association rule. A set of different linking rules is applied based on the language structure. Each connection rule creates a semantic connection relationship between the semantic objects. The method for applying the connection rule includes the following steps.

1. Relationships for parent and child pairs are identified.

2. Compatibility for the meaning category of the parent and child is checked. If the parent and child are semantically compatible, a semantic relationship between the parent and child is created.

3. If the parent-child compatibility is not established, corrective rules are applied to identify the appropriate parent-child meaning and create a semantic relationship between the parent and child.

4. The process continues until all pairs contain a semantic relationship.

All linking rules are used to specify the parent-child relationship and the semantic category of the parent and child. A score is assigned to a semantic connection relationship based on the compatibility. The following features are considered in the linking rules.

1. Distinction between different relationships, such as after, before, between, at, on, over.

2. Interpret cardinal numbers using different prepositions. For example, '9' in "at 9" is treated as an hour. The '9' in "on 9" is treated as a date. The above process becomes the basic interpretation in the absence of any evidence for other interpretations.

3. Extraction of referential expressions, quantifiers and repetitive words from temporal expressions.

4. Semantic representation and independent language creation.

5. Inference of temporal relationships without specific prepositions.

6. Infer and extend a common temporal expression through two unrelated temporal terms in the same sentence. For example, in the sentence "let's meet John on Monday and Tuesday at 6pm", 6pm (6pm) is inferred as a common time that applies to both dates. can be

In step 610, the operation 600 includes evaluating the accuracy of meaning. That is, in step 610, the LE unit 106 evaluates the accuracy of the meaning. After the semantic expressions are generated, the rules of semantics are used to check and score the correctness of the generated semantic expressions.

In step 612, the operation 600 includes generating a semantic inference. That is, in step 612, the LE unit 106 generates semantic inference. The semantic reasoning rule is applied for behavior transformation, tense information extraction, and intention extraction. For example, in the sentence "I will see John tomorrow", "see John" means "look at John" semantically. Rather, it means "meet John".

At step 614 , the operation 600 includes passing the scored semantic representations as output. That is, in step 614, the LE unit 106 delivers the scored semantic representations. When all the processing steps are completed, the scored semantic representation is transmitted to the input of the CE unit 108 .

7A shows an example for converting user input into a normative representation.

Referring to FIG. 7A , the reference time is June 10, 2015 at 2pm (June 10, 2015 at 2pm). 7A , the user of the electronic device 100 inputs text information (“John is going to Delhi from London 2nyt.)” in the memo. When the text information is input, the pre-processing unit 104 analyzes slang or spelling accuracy of the input text. The pre-processing unit 104 also converts the text information into a normative representation ("John is going to Delhi from London tonight"), as shown at the bottom of FIG. 7A . convert

7B shows an example of syntactic parsing.

Referring to FIG. 7B , the normative representation of the input text is analyzed by a syntax analyzer (not shown). The parser extracts different lexical items from the input text "John is going to Delhi from London tonight", and based on the relationship of the lexical items, the Connect vocabulary items. In addition, the syntax analyzer displays the input text in the form of a parse tree as shown in FIG. 7B .

7C shows an example of an operation for generating a semantic representation.

Referring to FIG. 7C , the LE unit 106 may extract information from the normative representation to generate semantic representations of the input text. The information may be an action, with whom, a source location, a destination location, a time, and the like. Also, as shown in FIG. 7C , the LE unit 106 uses a relationship between different lexical items and generates a semantic expression based on the extracted information.

7D illustrates an example of an operation for extracting a context.

Referring to FIG. 7D , the CE unit 108 receives the semantic expression from the LE unit 106, and distinguishes ambiguous sentences and phrases from the context. In addition, the CE unit 106 generates an appropriate semantic interpretation. For example, as shown in FIG. 7D , the CE unit 106 sets the time to 'June 10, 2015 8 PM in order to more clearly distinguish the meaning of 'tonight' in the received semantic expression. The hour (June 10 2015, 8:00pm)' can be changed to 'June 10, 2015 11:59pm (June 10 2015, 11:59pm)'.

7E shows another example of syntactic parsing.

Referring to FIG. 7E , the parser analyzes different lexical entries in the second sentence of the memo (“I will meet him there on the last Tuesday of next month”). extracts and connects the vocabulary items based on the relationship between the vocabulary items. In addition, the parser generates a parse tree for the input as shown in FIG. 7E .

7F shows another example of an operation for generating a semantic representation.

Referring to FIG. 7F , the LE unit 106 may extract information from the normative representation to generate semantic representations of the input text. In addition, the LE unit 106 uses the relationship between the different vocabulary items as shown in FIG. 7F , and generates a semantic expression based on the extracted information.

7G shows an example of the operation of merging semantic representations.

Referring to FIG. 7G , the CE unit 108 generates the multiple semantic expressions (ie, the two sentences "John is going to Delhi from London tonight" and "I will meet him there on the semantic expressions for "last Tuesday of next month").

7H shows another example of an operation for extracting a context.

Referring to FIG. 7H , the CE unit 108 extracts contextual information from the semantic representation merged by the process illustrated in FIG. 7G . The context information may mean information for determining the context. For example, the CE unit 108 may extract the type of “sender” corresponding to “who” in the operation reminder as human as shown in FIG. 7H . there is. In addition, as shown in FIG. 7H , the CE unit 108 extracts the type of “John” corresponding to “whom” in the operation reminder as human and the gender as male. can do.

7I shows another example of syntax parsing.

Referring to FIG. 7I , the parser is used for the sentence "We plan to have coffee at 7 in the evening at Starbucks that day." Create a parse tree. Also, the syntax analyzer extracts different vocabulary items and establishes a connection relationship based on the relationship between the vocabulary items.

7J shows another example of an operation for generating a semantic representation.

Referring to FIG. 7J , based on a dictionary analysis of text information, the LE unit 106 uses a relationship between different lexical items and generates a first level of semantic expression.

7K shows another example of the operation of merging semantic representations.

Referring to FIG. 7K , the CE unit 108 merges multiple semantic representations to generate an appropriate semantic representation.

7L shows another example of an operation for extracting a context.

Referring to FIG. 7L , the CE unit 108 extracts context information from the generated plurality of semantic representations.

7M illustrates an example of an operation of automatically converting a memo sent to me into an action reminder based on a semantic expression.

Referring to FIG. 7M , the CE unit 108 extracts context information from the generated semantic expression and distinguishes ambiguous sentences and phrases from the context to generate an appropriate reminder.

8 shows an example of an operation of automatically converting a memo sent to me into an action reminder.

Referring to FIG. 8 , the user of the electronic device 100 displays text information “I am going home on Thanksgiving weekend. I am going to home on Thanks Giving weekend. I want to catch” up with Susan then.)". The input text information will be converted into a reminder as shown in FIG. 7 .

9 shows another example of an operation for automatically converting a memo sent to me into an action reminder.

Referring to FIG. 9 , the user of the electronic device 100 inputs text information "Credit card payment is 4 days. Don't forget. (Credit Card (CC) bill is due on 4th. Don't forget it.)" . The input text information will be converted into a reminder as shown in FIG. 8 .

10 illustrates a computing environment 1002 implementing a method for automatically converting a memo sent to me by an electronic device into an action reminder.

Referring to FIG. 10 , the computing environment 1002 includes at least one processing unit 1008 , a memory 1010 , a storage unit 1012 , a plurality of network devices 1016 , and a plurality of input output (I/O) devices 1014 . . The at least one processing unit 1008 comprises a control unit 1004 and an arithmetic logic unit (ALU) 1006 . The processing unit 1008 is configured to process instructions of an algorithm. The processing unit 1008 receives an instruction from the control unit 1008 and performs processing of the instructions. In addition, logical and arithmetic operations on the instructions are performed by the ALU 1006.

Methods according to the embodiments described in the claims or specifications of the present disclosure may be implemented in the form of hardware, software, or a combination of hardware and software. When implemented in software, a computer-readable storage medium storing one or more programs (software modules) may be provided. One or more programs stored in the computer-readable storage medium are configured to be executable by one or more processors in an electronic device (device). One or more programs include instructions for causing an electronic device to execute methods according to embodiments described in a claim or specification of the present disclosure.

Such programs (software modules, software) include random access memory, non-volatile memory including flash memory, read only memory (ROM), electrically erasable programmable ROM (EEPROM: Electrically Erasable Programmable Read Only Memory), magnetic disc storage device, Compact Disc-ROM (CD-ROM), Digital Versatile Discs (DVDs), or any other form of It may be stored in an optical storage device or a magnetic cassette. Alternatively, it may be stored in a memory composed of a combination of some or all thereof. In addition, each configuration memory may be included in plurality.

In addition, the program is transmitted through a communication network composed of a communication network such as the Internet, Intranet, Local Area Network (LAN), Wide LAN (WLAN), or Storage Area Network (SAN), or a combination thereof. It may be stored on an attachable storage device that can be accessed. Such a storage device may be connected to a device implementing an embodiment of the present disclosure through an external port. In addition, a separate storage device on the communication network may be connected to the device implementing the embodiment of the present disclosure.

In the specific embodiments of the present disclosure described above, components included in the disclosure are expressed in the singular or plural according to the specific embodiments presented. However, the singular or plural expression is appropriately selected for the context presented for convenience of description, and the present disclosure is not limited to the singular or plural element, and even if the element is expressed in plural, it is composed of the singular or singular. Even an expressed component may be composed of a plurality of components.

Meanwhile, although specific embodiments have been described in the detailed description of the present disclosure, various modifications are possible without departing from the scope of the present disclosure. Therefore, the scope of the present disclosure should not be limited to the described embodiments and should be defined by the claims described below as well as the claims and equivalents.

Claims (10)

A method of operating an electronic device, comprising:
detecting an input including at least one vocabulary from a user of the electronic device;
converting each of the at least one vocabulary into a canonical representation by analyzing an anaphora expression or a deictic expression for each of the at least one vocabulary;
generating a plurality of semantic representations from each of the normative representations;
determining a context based on the plurality of semantic expressions; and
and generating a reminder based on the determined context.
The method according to claim 1, wherein the process of converting each of the at least one vocabulary into a normative representation comprises:
and analyzing the surrogate or referential expression based on at least one of time information, location information, participant information, purpose information, gender information, and number information for each of the at least one vocabulary.
delete The method according to claim 1, wherein the determining of the context based on the plurality of semantic expressions comprises:
determining a degree of agreement between at least two contexts for the input based on the analyzed surrogate or referential expression and a predetermined parameter; and
and determining the context by merging attribute values included in each of the at least two contexts based on the degree of matching.
The method of claim 4, wherein the merging of attribute values included in each of the at least two contexts comprises:
and performing at least one of attaching, overlapping, and invalidating the attribute values included in each of the at least two contexts.
In an electronic device,
at least one processor; and
a memory operatively coupled to the at least one processor;
The memory, when executed, the at least one processor,
detecting an input including at least one vocabulary from a user of the electronic device;
Analyze an anaphora expression or a deictic expression for each of the at least one vocabulary, and transform each of the at least one vocabulary into a canonical representation,
generate a plurality of semantic representations from each of the normative representations;
determine a context based on the plurality of semantic expressions;
An electronic device that stores instructions for generating a reminder based on the determined context.
7. The method of claim 6,
The instructions, the at least one processor,
The electronic device to analyze the surrogate expression or the referential expression based on at least one of time information, location information, participant information, purpose information, gender information, and number information for each of the at least one vocabulary.
delete 7. The method of claim 6,
The instructions, the at least one processor,
determining a degree of agreement between at least two contexts for the input based on the analyzed surrogate or referential expression and a predetermined parameter;
and to determine the context by merging attribute values included in each of the at least two contexts based on the degree of matching.
10. The method of claim 9,
The instructions, the at least one processor,
and performing at least one of an operation of attaching, overlapping, and invalidating the attribute values included in each of the at least two contexts.

Images