52 Matching Results

Search Results

Machine Language Techniques for Conversational Agents

Description: Machine Learning is the ability of a machine to perform better at a given task, using its previous experience. Various algorithms like decision trees, Bayesian learning, artificial neural networks and instance-based learning algorithms are used widely in machine learning systems. Current applications of machine learning include credit card fraud detection, customer service based on history of purchased products, games and many more. The application of machine learning techniques to natural language processing (NLP) has increased tremendously in recent years. Examples are handwriting recognition and speech recognition. The problem we tackle in this Problem in Lieu of Thesis is applying machine-learning techniques to improve the performance of a conversational agent. The OpenMind repository of common sense, in the form of question-answer pairs is treated as the training data for the machine learning system. WordNet is interfaced with to capture important semantic and syntactic information about the words in the sentences. Further, k-closest neighbors algorithm, an instance based learning algorithm is used to simulate a case based learning system. The resulting system is expected to be able to answer new queries with knowledge gained from the training data it was fed with.
Date: December 2003
Creator: Sule, Manisha D.
Partner: UNT Libraries

Networks and Natural Language Processing

Description: Article discussing networks and natural language processing. The authors present some of the most successful graph-based representations and algorithms used in language processing and try to explain how and why they work.
Date: September 2008
Creator: Radev, Dragomir R. & Mihalcea, Rada, 1974-
Partner: UNT College of Engineering

UNT 2005 TREC QA Participation: Using Lemur as IR Search Engine

Description: This paper reports the authors' TREC 2005 QA participation. The authors' QA system Eagle QA developed last year was expanded and modified for this year's QA experiments. Particularly, the authors used Lemur 4.1 as the Information Retrieval (IR) Engine this year to find documents that may contain answers for the test questions from the document collection. The authors' result shows Lemur did a reasonable job on finding relevant documents. But certainly there is room for further improvement.
Date: 2005
Creator: Chen, Jiangping; Yu, Ping & Ge, He
Partner: UNT College of Information

Text Semantic Similarity, with Applications

Description: In this paper, the authors present a knowledge-based method for measuring the semantic-similarity of texts. Through experiments performed on two different applications: (1) paraphrase and entailment identification, and (2) word sense similarity, the authors show that this method outperforms the traditional text similarity metrics based on lexical matching.
Date: September 2005
Creator: Corley, Courtney; Csomai, Andras & Mihalcea, Rada, 1974-
Partner: UNT College of Engineering

SenseLearner: Minimally Supervised Word Sense Disambiguation for All Words in Open Text

Description: This paper introduces SenseLearner - a minimally supervised sense tagger that attempts to disambiguate all content words in a text using the sense from WordNet. SenseLearner participated in the SENSEVAL-3 English all words task, and achieved an average accuracy of 64.6%.
Date: 2004
Creator: Mihalcea, Rada, 1974- & Faruque, Ehsanul
Partner: UNT College of Engineering

Unsupervised Large-Vocabulary Word Sense Disambiguation with Graph-based Algorithms for Sequence Data Labeling

Description: This paper introduces a graph-based algorithm for sequence data labeling, using random walks on graphs encoding label dependencies. The algorithm is illustrated and tested in the context of an unsupervised word sense disambiguation problem, and shown to significantly outperform the accuracy achieved through individual label assignment, as measured on standard sense-annotated data sets.
Date: October 2005
Creator: Mihalcea, Rada, 1974-
Partner: UNT College of Engineering

Finding Semantic Associations on Express Lane

Description: This paper introduces a new codification scheme for efficient computation of measures in semantic networks. The scheme is particularly useful for fast computation of semantic associations between words and implementation of an informational retrieval operator for efficient search in semantic spaces. Other applications may also be possible.
Date: May 2004
Creator: Nastase, Vivi & Mihalcea, Rada, 1974-
Partner: UNT College of Engineering