Wlodek Zadrozny

Language model adaptation via network of similar users

A language recognition system, method and program product for recognizing language based input from computer users on a network of connected computers. Each computer includes at least one user based language model trained for a corresponding user for automatic speech recognition, handwriting recognition, machine translation, gesture recognition or other similar actions that require interpretation of user activities. Network computer users are clustered into classes of similar users according to user similarities such as, nationality, profession, sex, age, etc. User characteristics are collected by sensors and from databases and, then, distributed over the network during user activities. Language models with similarities among similar users on the network are identified. The language models include a language model domain, with similar language models being clustered according to their domains. Language models identified as similar are modified in response to user production activities. After modification of one language model, other identified similar language models are compared and adapted. Also, user data, including information about user activities and language model data, is transmitted over the network to other similar users. Language models are adapted only in response to similar user activities, when these activities are recorded and transmitted over the network. Language models are given a global context based on similar users that are connected together over the network.

From compositional to systematic semantics

We prove a theorem stating that any semantics can be encoded as a compositional semanties, which means that, essentially, the standard definition of compositionality is formally vacuous. We then show that when compositional semantics is required to be “systematic” (that is, the meaning function cannot be arbitrary, but must belong to some class), it is possible to distinguish between compositional and noncompositional semantics. As a result, we believe that the paper clarifies the concept of compositionality and opens the possibility of making systematic formal comparisons of different systems of grammar.

Method and apparatus for creating speech recognition grammars constrained by counter examples

An automated system generates and revises grammars for speech recognizers in a speech recognition system. Given an initial grammar, expressed in terms of non-terminals in Backus-Naur Form (BNF) notation, a sentence generator generates a list of all sentences accepted by the grammar. From this list, a corpus of inappropriate or irrelevant sentences which are accepted by the grammar (counter-examples) is identified. A grammar revisor program uses the original grammar and the list of counter examples, to generate a pruned list from which a revised grammar is generated. The revision process is iterated several times either concatenating or merging pairs of non-terminals until the revised grammar is deemed satisfactory in that it accepts as legal only relevant sentences. The revised grammar is used by the speech recognizer, thus reducing errors in the overall system.

Natural Language Assistant: A Dialog System for Online Product Recommendation

With the emergence of electronic-commerce systems, successful information access on electroniccommerce web sites becomes essential. Menu-driven navigation and keyword search currently provided by most commercial sites have considerable limitations because they tend to overwhelm and frustrate users with lengthy, rigid, and ineffective interactions. To provide an efficient solution for information access, we have built the NATURAL language ASSISTANT (NLA), a web-based natural language dialog system to help users find relevant products on electronic-commerce sites. The system brings together technologies in natural language processing and human-computer interaction to create a faster and more intuitive way of interacting with web sites. By combining statistical parsing techniques with traditional AI rule-based technology, we have created a dialog system that accommodates both customer needs and business requirements. The system is currently embedded in an application for recommending laptops and was deployed as a pilot on IBMs web site.

Textual resource acquisition and engineering

A key requirement for high-performing question-answering (QA) systems is access to high-quality reference corpora from which answers to questions can be hypothesized and evaluated. However, the topic of source acquisition and engineering has received very little attention so far. This is because most existing systems were developed under organized evaluation efforts that included reference corpora as part of the task specification. The task of answering Jeopardy!™ questions, on the other hand, does not come with such a well-circumscribed set of relevant resources. Therefore, it became part of the IBM Watson™ effort to develop a set of well-defined procedures to acquire high-quality resources that can effectively support a high-performing QA system. To this end, we developed three procedures, i.e., source acquisition, source transformation, and source expansion. Source acquisition is an iterative development process of acquiring new collections to cover salient topics deemed to be gaps in existing resources based on principled error analysis. Source transformation refers to the process in which information is extracted from existing sources, either as a whole or in part, and is represented in a form that the system can most easily use. Finally, source expansion attempts to increase the coverage in the content of each known topic by adding new information as well as lexical and syntactic variations of existing information extracted from external large collections. In this paper, we discuss the methodology that we developed for IBM Watson for performing acquisition, transformation, and expansion of textual resources. We demonstrate the effectiveness of each technique through its impact on candidate recall and on end-to-end QA performance.

Natural language dialogue for personalized interaction

T he pragmatic goal of natural language (NL) and multimodal interfaces (speech recognition, keyboard entry, pointing, among others) is to enable ease-of-use for users/customers in performing more sophisticated human-computer interactions (HCI). NL research attempts to define extensive discourse models that in turn provide improved models of context-enabling HCI and personalization. Customers have the initiative to Technologies that successfully recognize and react to spoken or typed words are key to true personalization. Front-and back-end systems must respond in accord, and one solution may be found somewhere in the middle(ware).

Statistical source expansion for question answering

A source expansion algorithm automatically extends a given text corpus with related content from large external sources such as the Web. The expanded corpus is not intended for human consumption but can be used in question answering (QA) and other information retrieval or extraction tasks to find more relevant information and supporting evidence. We propose an algorithm that extends a corpus of seed documents with web content, using a statistical model to select text passages that are both relevant to the topics of the seeds and complement existing information. In an evaluation on 1,500 hand-labeled web pages, our algorithm ranked text passages by relevance with 81% MAP, compared to 43% when relying on web search engine ranks alone and 75% when using a multi-document summarization algorithm. Applied to QA, the proposed method yields consistent and significant performance gains. We evaluated the impact of source expansion on over 6,000 questions from the Jeopardy! quiz show and TREC evaluations using Watson, a state-of-the-art QA system. Accuracy increased from 66% to 71% on Jeopardy! questions and from 59% to 64% on TREC questions.

The Role of a Natural Language Conversational Interface in Online Sales: A Case Study

This paper describes the evaluation of a natural language dialog-based navigation system (HappyAssistant) that helps users access e-commerce sites to find relevant information about products and services. The prototype system leverages technologies in natural language processing and human-computer interaction to create a faster and more intuitive way of interacting with websites, especially for less experienced users. The result of a comparative study shows that users prefer the natural language-enabled navigation two to one over the menu driven navigation. In addition, the study confirmed the efficiency of using natural language dialog in terms of the number of clicks and the amount of time required to obtain the relevant information. In the case study, as compared to the menu driven system, the average number of clicks used in the natural language system was reduced by 63.2% and the average time was reduced by 33.3%.

Evolution of the conversation machine: a case study of bringing advanced technology to the marketplace

This paper describes the evolution of the Conversation Machine, a conversational speech system which allows users to carry out common banking transactions over the telephone using a conversational-style interface. The discussion is organized according to three phases of the project--the demonstration, laboratory, and customer phases. The different phases of the project had different goals and brought different design issues to the for&ont. In particular, the realities of worl&g with a customer partner have affected the design of the user interface and functionality of the system in ways not anticipated at earlier stages of the project.

On rules of abduction

We present a logical theory of abduction based on the idea of recognizing explanation, or abduction, as a separate reasoning activity. We describe a formalism for writing rules of abduction; furthermore, we define a validity criterion for such rules. The criterion is based on the concept of invariants. This idea allows us to link abduction with induction and deduction. We believe that the three types of inference rules can best be understood in terms of symmetry, i.e. types of relations they preserve, namely: explainability, falsifiability and truth. We also formulate a model theory of abduction and link it with a proof theory. We discuss a variety of rules of abduction and argue that logical forms of abduction do not have to be restricted to the reversemodus ponens. These rules are used to describe such tasks as word-sense disambiguation and anaphora resolution in natural language processing, as well as abduction-based diagnosis.