翻译 | Placing Search in Context The Concept Revisited

原文

摘要

[1] Keyword-based search engines are in widespread use today as a popular means for Web-based information retrieval.

[2] Although such systems seem deceptively simple, a considerable amount of skill is required in order to satisfy non-trivial information needs.

[3] This paper presents a new conceptual paradigm for performing search in context, that largely automates the search process, providing even non-professional users with highly relevant results.

[4] This paradigm is implemented in practice in the IntelliZap system, where search is initiated from a text query marked by the user in a document she views, and is guided by the text surrounding the marked query in that document (“the context”).

[5] The context-driven information retrieval process involves semantic keyword extraction and clustering to automatically generate new, augmented queries.

[6] The latter are submitted to a host of general and domain-specific search engines.

[7] Search results are then semantically reranked, using context. Experimental results testify that using context to guide search, effectively offers even inexperienced users an advanced search tool on the Web.

模型改进

第一节

[1] The core of IntelliZap technology is a semantic network, which provides a metric for measuring distances between pairs of words.

[2] The basic semantic network is implemented using a vector-based approach, where each word is represented as a vector in multi-dimensional space.

[3] To assign each word a vector representation, we first identified 27 knowledge domains (such as computers, business and entertainment) roughly partitioning the whole variety of topics.

[4] We then sampled a large set of documents in these domains on the Internet Word vectors were obtained by recording the frequencies of each word in each knowledge domain.

[5] Each domain can therefore be viewed as an axis in the multi-dimensional space.

[6] The distance measure between word vectors is computed using a correlation-based metric:

第二节

[1] Unfortunately, there are no accepted procedures for evaluating performance of semantic metrics.

[2] Following Resnik [1999], we evaluated different metrics by computing correlation between their scores and human-assigned scores for a list of word pairs.

[3] The intuition behind this approach is that a good metric should approximate human judgments well.

[4] While Resnik used a list of 30 noun pairs from Miller and Charles [1991], we opted for a more comprehensive evaluation.

[5] To this end, we prepared a diverse list of 350 noun pairs representing various degrees of similarity,10 and employed 16 subjects to estimate the “relatedness” of the words in pairs on a scale from 0 (totally unrelated words) to 10 (very much related or identical words).

[6] The vector-based metric achieved 41% correlation with averaged human scores, and the WordNet-based metric achieved 39% correlation11,12 A linear combination of the two metrics achieved 55% correlation with human scores.

[7] Currently, our semantic network is defined for the English language, though the technology can be adapted for other languages with minimal effort.

[8] This would require training the network using textual data for the desired language, properly partitioned into domains.

[9] Linguistic information can be added, subject to the availability of adequate tools for the target language (e.g., EuroWordNet for European languages [Euro WordNet] or EDR for Japanese [Yokoi 1995]).

翻译

摘要

[1] 基于关键字的搜索引擎作为一种流行的基于Web的信息检索手段，在今天得到了广泛的应用。

[2] 虽然这样的系统看起来似乎很简单，但为了满足非琐碎的信息需求，需要大量的技巧。

[3] 本文提出了一种新的在上下文中执行搜索的概念范式，它在很大程度上自动化了搜索过程，甚至为非专业用户召回了高度相关的结果。

[4] 这种范例是在 Intellizap 系统中实现的。在该系统中，搜索从用户在其所查看的文档中标记的文本查询开始，并由该文档中标记的查询周围的文本（“上下文”）来引导。

[5] 上下文驱动的信息检索过程包括语义关键字提取和聚类，从而自动生成新的、扩充的查询。

[6] 后者被提交给一系列通用和特定于域的搜索引擎。

[7] 然后使用上下文对搜索结果进行语义重新排序。实验结果表明，利用上下文来引导搜索，甚至可以有效地为没有经验的用户提供一种先进的网络搜索工具。

模型改进

第一节

[1] Intellizap技术的核心是一个语义网络，它为测量成对词之间的距离提供了一个度量标准。

[2] 基本语义网络是使用基于向量的方法实现的，其中每个词在多维空间中表示为一个向量。

[3] 为了给每个单词分配一个向量表示，我们首先确定了27个知识域（如计算机、商业和娱乐），大致划分了各种主题。

[4] 然后，我们对这些领域中的大量文档进行了抽样，通过记录每个知识领域中每个单词的频率，获得了互联网上的单词向量。

[5] 因此，可以将每个域看作多维空间中的一个轴。

[6] 单词向量之间的距离度量是使用基于相关性的度量来计算的：

第二节

[1] 不幸的是，没有可以被接受的手段来评估语义度量的性能。

[2] 继 Resnik[1999] 之后，我们通过计算机器打分与人类对指定的单词打分列表之间的相关性，来评估不同的指标。

[3] 这种方法背后的直觉是，一个好的度量应该很好地近似人类的判断。

[4] 虽然 Resnik 使用了 Miller 和 Charles[1991] 的 30 个名词对列表，但我们选择了更全面的评估。

[5] 为此，我们准备了一份 350 个不同的名词词对的列表，分别代表不同程度的相似性，由 10 个和 16 个受试者，以从0（完全无关的词）到10（非常相关或相同的词）的尺度来估计词对间的“相关性”。

[6] 基于向量的度量与平均人类分数的相关性达到41%，基于 WordNet 的度量与平均人类分数的相关性达到 39%，11,12这两个度量的线性组合与人类分数的相关性达到55%。

[7] 目前，我们的语义网络是为英语定义的，尽管这项技术可以用最少的努力适应其他语言。

[8] 这需要使用目标语言的文本数据对网络进行培训，并将其正确划分为域。

[9] 可根据目标语言的适当工具（例如，欧洲语言的 EurowordNet [欧元wordNet] 或日语的 EDR[Yokoi 1995]）添加语言信息。