Xerox Corporation

Naver Corporation

Conduent Business Services, LLC

Caroline Brun

Caroline Hagege

Marc Dymetman

Claude Roux

Caroline Hagège

Salah Aït-Mokhtar

Nikolaos Lagos

Agnes Sandor

Hervé Déjean

Stefania Castellani

Frederic Roulland

Julien Perez

Antonietta Grasso

Vassilina Nikoulina

Frederique Segond

Guillaume Jacquet

Fr&eacute;d&eacute;rique Segond

Diana Nicoleta Popa

Ngoc Phuoc An Vo

Maud Ehrmann

Aurelien Max

Roman Castagne

Ágnes Sandor

Anna Stavrianou

An annotation assistant for semi-automatic creation of annotated machine learning training data comprises a graphical annotation interface provider that provides a graphical annotation interface for displaying a sequence of tokens to be annotated and allowing a span selection by a user. An automatic span tagger comprises an artificial neural network configured to assign class probabilities to the tokens of the sequence of tokens and a processor-based path-scoring algorithm for computing scores for semantic attribute paths of the span selection. The graphical annotation interface is further configured for displaying the scored semantic attribute paths, and for receiving a selection among the scored semantic attribute paths from the user.

Method and system for extraction and annotation using semantic attribute paths

A computer-implemented method for training a neural end-to-end aspect based sentiment analysis (ABSA) system includes: inputting a batch of samples of a dataset into the neural end-to-end ABSA system, where the neural end-to-end ABSA system includes: a contextual language encoder configured to embed tokens with context; a first self-attention network configured to, based on an output of the contextual language encoder, detect an aspect term and provide a first output corresponding to the aspect term; and a second self-attention network configured to, based on the output of the contextual language encoder, detect the aspect term and provide a second output corresponding to the aspect term; and based on the inputted batch of samples and a consistency loss function, selectively adjusting weights of the neural end-to-end ABSA system based on consistent aspect term detection by the first self-attention network and the second self-attention network.

Multilingual, end-to-end-aspect based sentiment analysis with opinion triplets predictions

In methods and systems for providing information to a user relating to a point-of-interest (POI), a query is received via a communication network. A processor receives information relating to at least one POI based on the received query. The retrieved information comprises a set of retrieved POI tips relating to the POI. Each of the set of retrieved POI tips is previously extracted from POI free-text reviews and respectively associated in the data storage with one or more semantic categories based on a predefined tip ontology. The processor generates a user interface interactively displaying the retrieved POI tips and the semantic categories associated with each of the POI tips. The displayed POI tips are sorted by their associated displayed semantic categories. The generated user interface is transmitted to a computing device for display.

Method and system for providing information to a user relating to a point-of-interest

A method for categorizing an issue includes, for each of a plurality of categories of issue, providing at least one discourse pattern for identify text sequences that meet the discourse pattern. At least one of the discourse patterns specifies that an instance of a domain term in a domain term vocabulary be present in the text sequence for the pattern to be met. An issue is received which includes a text sequence. The text sequence is categorized based on which, if any, of the discourse patterns are met by the text sequence of the received issue. Information based on the categorization of the text sequence is output.

Method and system for identifying user issues in forum posts based on discourse analysis

Systems and methods for aspect-based opinion mining including identifying the polarity (e.g., positive, negative, etc.) of different features of a product or a service as expressed in a text. This general task can be divided into four sub-tasks: identifying the aspect terms, classifying them into one of a set of predefined aspect categories, and identifying the polarity of the aspects terms and the aspect categories. A combination of systems (e.g., rule-based and machine learning based) can be employed to implement aspect category and aspect term polarity mutual annotation for aspect-based sentiment analysis.

Category and term polarity mutual annotation for aspect-based sentiment analysis

A method for aspect categorization includes receiving an input text sequence and identifying aspect terms and sentiment phrases in the input text sequence, where present. For an identified aspect term, identifying sentiment dependencies in which the aspect term is in a syntactic dependency with one of the identified sentiment phrases, and identifying pseudo-dependencies from a dependency graph of the input text sequence. The dependency graph includes a sequence of nodes. In a pseudo-dependency, a node representing the aspect term precedes or follows a node representing a semantic anchor in the dependency graph without an intervening other aspect term. Features for the aspect term are extracted from at least one of identified sentiment dependencies and identified pseudo-dependencies. With a classifier trained to output at least one of category labels and polarity labels for aspect terms, classifying the identified aspect term based on the extracted features.

Loose term-centric representation for term classification in aspect-based sentiment analysis

Systems and methods for acquiring information associated with a user by using NLP techniques are disclosed. One or more phrases are classified in one or more categories at least partly on the basis of a period for which a product has been used by the user, the user's experience with the product, preferences of the user, or needs of the user by applying one or more natural language processing (NLP) techniques. The one or more phrases are extractable from an electronic publication at least partly on the basis of on a predefined set of verbs, a predefined set of domain-specific terms, and terms indicative of temporal information. One or more terms from the classified phrases are extracted, in which the one or more terms are indicative of the information about the user.

Methods and systems for acquiring user related information using natural language processing techniques

A system and method for natural language processing of queries are provided. A lexicon includes text elements that are recognized as being a proper noun when capitalized. A natural language query includes a sequence of text elements including words. The query is processed. The processing includes a preprocessing step, in which part of speech features are assigned to the text elements in the query. This includes identifying, from a lexicon, a text element in the query which starts with a lowercase letter and assigning recapitalization information to the text element in the query, based on the lexicon. This information includes a part of speech feature of the capitalized form of the text element. Then parts of speech for the text elements in the query are disambiguated, which includes applying rules for recapitalizing text elements based on the recapitalization information.

Linguistically-adapted structural query annotation

A system and method for extraction of suggestions for improvement form a corpus of documents, such as customer reviews, are disclosed. A structured terminology provided or a topic includes a set of semantic classes, each including a set of terms. A thesaurus of terms relating to suggestions of improvement is provided. Text elements of text strings in the documents which are instances of terms in the structured terminology are labeled with the corresponding semantic class and text elements which are instances of terms in the thesaurus are also labeled. A set of patterns is applied to the labeled text strings to identify suggestions of improvement expressions. The patterns define syntactic relations between text elements, some of which are required to be instances of one of the terms in a particular semantic class or thesaurus. A set of suggestions for improvements is output based on the identified suggestions of improvement expressions.

System and method for suggestion mining

An authoring assistant includes a parser which automatically identifies opinion expressions in input text. The text may include an author's review of an item, such as a product or service. A computer-implemented opinion review component generates an analysis of the text, which is based on the identified opinion expressions. The opinion review component computes an effective opinion of the text as a function of a measure of polarity associated with the identified opinion expressions. A representation generator generates a representation of the analysis for display on an associated user interface. The representation of the analysis includes a representation of the effective opinion. In the case of a review, the authoring assistant may allow the author to modify the review to reduce incoherence with a rating of the item.

Customer review authoring assistant

A system, method, and computer program product for generating a polar vocabulary are provided. The method includes extracting textual content from each review in a corpus of reviews. Each of the reviews includes an author's rating, e.g., of a specific product or service to which the textual content relates. A set of frequent nouns is identified from the textual content of the reviews. Adjectival terms are extracted from the textual content of the reviews. Each adjectival term is associated in the textual content with one of the frequent nouns. A polar vocabulary including at least some of the extracted adjectival terms is generated. A polarity measure is associated with each adjectival term in the vocabulary which is based on the ratings of those reviews from which the adjectival term was extracted.

Corpus-based system and method for acquiring polar adjectives

A computer-implemented system and method are provided for warning a user of a missing attachment to an email. The method may include automatically recognizing a natural language of text of an email and selecting a keyword list from a plurality of keyword lists, based on the recognized natural language. Each keyword list is associated with a respective natural language and includes at least one keyword. At least one of the keyword lists includes a multi-sense keyword having a plurality of senses. A first of the plurality of senses is recognized as referring to an attachment and a second of the plurality of senses is recognized as not referring to an attachment. The text of the email is processed to identify an instance, where present, of a keyword that is in the selected keyword list and, for a keyword which is a multi-sense keyword, at least one sense-related rule is applied to a portion of the text which includes the instance of the multi-sense keyword. Based on the application of the at least one sense-related rule, where the email lacks an attachment, a notification is provided to the user.

Linguistically enhanced email detector

A method for named entity resolution includes parsing an input text string to identify a context in which an identified named entity of the input text string is used. The identified context is compared with at least one stored context in which the named entity in the stored context is associated with a class of named entity, the named entity class being selected from a plurality of classes, at least one of the plurality of classes corresponding to a metonymic use of a respective named entity. A named entity class is assigned to the identified named entity from the plurality of named entity classes, based on at least one of the identified context and the comparison.

Hybrid system for named entity resolution

A system and method of developing rules for text processing enable retrieval of instances of named entities in a predetermined semantic relation (such as the DATE and PLACE of an EVENT) by extracting patterns from text strings in which attested examples of named entities satisfying the semantic relation occur. The patterns are generalized to form rules which can be added to the existing rules of a syntactic parser and subsequently applied to text to find candidate instances of other named entities in the predetermined semantic relation.

Semantically-driven extraction of relations between named entities

A system and method are provided for refining a user's query. An entity index, generated from a corpus of text documents, is provided. The entity index includes a set of entity structures, each including a plurality of terms. Each of the terms of an entity structure is a feature of the same entity. Entity structures can be retrieved from the entity index which match at least a portion of the user's query. Clusters of the retrieved entity structures are identified which have at least one of their terms in common. A cluster hierarchy is generated from the identified clusters in which nodes of the hierarchy are defined by one or more of the terms of the retrieved entity structures. At least a portion of the cluster hierarchy is presented to the user for facilitating refinement of the user's query through user selection of a node which, when formulated as a search, retrieves one or more responsive documents from the corpus of documents.

Growing community of inventors

Grenoble, France

Caroline Brun