Anagha Kulkarni

Name discrimination by clustering similar contexts

It is relatively common for different people or organizations to share the same name. Given the increasing amount of information available online, this results in the ever growing possibility of finding misleading or incorrect information due to confusion caused by an ambiguous name. This paper presents an unsupervised approach that resolves name ambiguity by clustering the instances of a given name into groups, each of which is associated with a distinct underlying entity. The features we employ to represent the context of an ambiguous name are statistically significant bigrams that occur in the same context as the ambiguous name. From these features we create a co–occurrence matrix where the rows and columns represent the first and second words in bigrams, and the cells contain their log–likelihood scores. Then we represent each of the contexts in which an ambiguous name appears with a second order context vector. This is created by taking the average of the vectors from the co–occurrence matrix associated with the words that make up each context. This creates a high dimensional “instance by word” matrix that is reduced to its most significant dimensions by Singular Value Decomposition (SVD). The different “meanings” of a name are discriminated by clustering these second order context vectors with the method of Repeated Bisections. We evaluate this approach by conflating pairs of names found in a large corpus of text to create ambiguous pseudo-names. We find that our method is significantly more accurate than the majority classifier, and that the best results are obtained by having a small amount of local context to represent the instance, along with a larger amount of context for identifying features, or vice versa.

Understanding temporal query dynamics

Web search is strongly influenced by time. The queries people issue change over time, with some queries occasionally spiking in popularity (e.g., earthquake) and others remaining relatively constant (e.g., youtube). The documents indexed by the search engine also change, with some documents always being about a particular query (e.g., the Wikipedia page on earthquakes is about the query earthquake) and others being about the query only at a particular point in time (e.g., the New York Times is only about earthquakes following a major seismic activity). The relationship between documents and queries can also change as peoples intent changes (e.g., people sought different content for the query earthquake before the Haitian earthquake than they did after). In this paper, we explore how queries, their associated documents, and the query intent change over the course of 10 weeks by analyzing query log data, a daily Web crawl, and periodic human relevance judgments. We identify several interesting features by which changes to query popularity can be classified, and show that presence of these features, when accompanied by changes in result content, can be a good indicator of change in query intent.

Automatic Cluster Stopping with Criterion Functions and the Gap Statistic

SenseClusters is a freely available system that clusters similar contexts. It can be applied to a wide range of problems, although here we focus on word sense and name discrimination. It supports several different measures for automatically determining the number of clusters in which a collection of contexts should be grouped. These can be used to discover the number of senses in which a word is used in a large corpus of text, or the number of entities that share the same name. There are three measures based on clustering criterion functions, and another on the Gap Statistic.

Document allocation policies for selective searching of distributed indexes

Indexes for large collections are often divided into shards that are distributed across multiple computers and searched in parallel to provide rapid interactive search. Typically, all index shards are searched for each query. For organizations with modest computational resources the high query processing cost incurred in this exhaustive search setup can be a deterrent to working with large collections. This paper investigates document allocation policies that permit searching only a few shards for each query (selective search) without sacrificing search accuracy. Random, source-based and topic-based document-to-shard allocation policies are studied in the context of selective search. A thorough study of the tradeoff between search cost and search accuracy in a sharded index environment is performed using three large TREC collections. The experimental results demonstrate that selective search using topic-based shards cuts the search cost to less than 1/5th of that of the exhaustive search without reducing search accuracy across all the three datasets. Stability analysis shows that 90% of the queries do as well or improve with selective search. An overlap-based evaluation with an additional 1000 queries for each dataset tests and confirms the conclusions drawn using the smaller TREC query sets.

Unsupervised Discrimination of Person Names in Web Contexts

Ambiguous person names are a problem in many forms of written text, including that which is found on the Web. In this paper we explore the use of unsupervised clustering techniques to discriminate among entities named in Web pages. We examine three main issues via an extensive experimental study. First, the effect of using a held---out set of training data for feature selection versus using the data in which the ambiguous names occur. Second, the impact of using different measures of association for identifying lexical features. Third, the success of different cluster stopping measures that automatically determine the number of clusters in the data.

An unsupervised language independent method of name discrimination using second order co-occurrence features

Previous work by Pedersen, Purandare and Kulkarni (2005) has resulted in an unsupervised method of name discrimination that represents the context in which an ambiguous name occurs using second order co–occurrence features. These contexts are then clustered in order to identify which are associated with different underlying named entities. It also extracts descriptive and discriminating bigrams from each of the discovered clusters in order to serve as identifying labels. These methods have been shown to perform well with English text, although we believe them to be language independent since they rely on lexical features and use no syntactic features or external knowledge sources. In this paper we apply this methodology in exactly the same way to Bulgarian, English, Romanian, and Spanish corpora. We find that it attains discrimination accuracy that is consistently well above that of a majority classifier, thus providing support for the hypothesis that the method is language independent.

SenseClusters: Unsupervised Clustering and Labeling of Similar Contexts

SenseClusters is a freely available system that identifies similar contexts in text. It relies on lexical features to build first and second order representations of contexts, which are then clustered using unsupervised methods. It was originally developed to discriminate among contexts centered around a given target word, but can now be applied more generally. It also supports methods that create descriptive and discriminating labels for the discovered clusters.

Selecting the right number of senses based on clustering criterion functions

This paper describes an unsupervised knowledge-lean methodology for automatically determining the number of senses in which an ambiguous word is used in a large corpus. It is based on the use of global criterion functions that assess the quality of a clustering solution.

Selective Search: Efficient and Effective Search of Large Textual Collections

The traditional search solution for large collections divides the collection into subsets (shards), and processes the query against all shards in parallel (exhaustive search). The search cost and the computational requirements of this approach are often prohibitively high for organizations with few computational resources. This article investigates and extends an alternative: selective search, an approach that partitions the dataset based on document similarity to obtain topic-based shards, and searches only a few shards that are estimated to contain relevant documents for the query. We propose shard creation techniques that are scalable, efficient, self-reliant, and create topic-based shards with low variance in size, and high density of relevant documents. The experimental results demonstrate that the effectiveness of selective search is on par with that of exhaustive search, and the corresponding search costs are substantially lower with the former. Also, the majority of the queries perform as well or better with selective search. An oracle experiment that uses optimal shard ranking for a query indicates that selective search can outperform the effectiveness of exhaustive search. Comparison with a query optimization technique shows higher improvements in efficiency with selective search. The overall best efficiency is achieved when the two techniques are combined in an optimized selective search approach.

Word Sense Disambiguation for Vocabulary Learning

Words with multiple meanings are a phenomenon inherent to any natural language. In this work, we study the effects of such lexical ambiguities on second language vocabulary learning. We demonstrate that machine learning algorithms for word sense disambiguation can induce classifiers that exhibit high accuracy at the task of disambiguating homonyms (words with multiple distinct meanings). Results from a user study that compared two versions of a vocabulary tutoring system, one that applied word sense disambiguation to support learning and another that did not, support rejection of the null hypothesis that learning outcomes with and without word sense disambiguation are equivalent, with a p-value of 0.001. To our knowledge this is the first work that investigates the efficacy of word sense disambiguation for facilitating second language vocabulary learning.