As Large Language Models (LLMs) rise in popularity, it is necessary to assess their capability in critically relevant domains. We present a comprehensive evaluation framework, grounded in science communication research, to assess LLM responses to questions about climate change. Our framework emphasizes both presentational and epistemological adequacy, offering a fine-grained analysis of LLM generations spanning 8 dimensions and 30 issues. Our evaluation task is a real-world example of a growing number of challenging problems where AI can complement and lift human performance. We introduce a novel protocol for scalable oversight that relies on AI Assistance and raters with relevant education. We evaluate several recent LLMs on a set of diverse climate questions. Our results point to a significant gap between surface and epistemological qualities of LLMs in the realm of climate communication.

Leonard Adolphs, Michelle Chen Huebscher, Christian Buck, Sertan Girgin, Olivier Bachem, Massimiliano Ciaramita, Thomas Hofmann. Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing. 2022.

We study the problem of linking the terms of a web-search query to a semantic representation given by the set of entities (a.k.a. concepts) mentioned in it. We introduce SMAPH, a system that performs this task using the information coming from a web search engine, an approach we call “piggybacking.” We employ search engines to alleviate the noise and irregularities that characterize the language of queries. Snippets returned as search results also provide a context for the query that makes it easier to disambiguate the meaning of the query. From the search results, SMAPH builds a set of candidate entities with high coverage. This set is filtered by linking back the candidate entities to the terms occurring in the input query, ensuring high precision. A greedy disambiguation algorithm performs this filtering; it maximizes the coherence of the solution by iteratively discovering the pertinent entities mentioned in the query. We propose three versions of SMAPH that outperform state-of-the-art solutions on the known benchmarks and on the GERDAQ dataset, a novel dataset that we have built specifically for this problem via crowd-sourcing and that we make publicly available.

In this paper we study the problem of linking open-domain web-search queries towards entities drawn from the full entity inventory of Wikipedia articles. We introduce SMAPH-2, a second-order approach that, by piggybacking on a web search engine, alleviates the noise and irregularities that characterize the language of queries and puts queries in a larger context in which it is easier to make sense of them. The key algorithmic idea underlying SMAPH-2 is to first discover a candidate set of entities and then link-back those entities to their mentions occurring in the input query. This allows us to confine the possible concepts pertinent to the query to only the ones really mentioned in it. The link-back is implemented via a collective disambiguation step based upon a supervised ranking model that makes one joint prediction for the annotation of the complete query optimizing directly the F1 measure. We evaluate both known features, such as word embeddings and semantic relatedness among entities, and several novel features such as an approximate distance between mentions and entities (which can handle spelling errors). We demonstrate that SMAPH-2 achieves state-of-the-art performance on the [email protected] benchmark. We also publish GERDAQ (General Entity Recognition, Disambiguation and Annotation in Queries), a novel, public dataset built specifically for web-query entity linking via a crowdsourcing effort. SMAPH-2 outperforms the benchmarks by comparable margins also on GERDAQ.

Zhendong Zhao, Lan Du, Benjamin Börschinger, John K Pate, Massimiliano Ciaramita, Mark Steedman, Mark Johnson. Proceedings of the 53rd Annual Meeting of the Association for Computational Linguistics and the 7th International Joint Conference on Natural Language Processing (Volume 1: Long Papers). 2015.

In this paper we design and implement a benchmarking framework for fair and exhaustive comparison of entity-annotation systems. The framework is based upon the definition of a set of problems related to the entity-annotation task, a set of measures to evaluate systems performance, and a systematic comparative evaluation involving all publicly available datasets, containing texts of various types such as news, tweets and Web pages. Our framework is easily-extensible with novel entity annotators, datasets and evaluation measures for comparing systems, and it has been released to the public as open source. We use this framework to perform the first extensive comparison among all available entity annotators over all available datasets, and draw many interesting conclusions upon their efficiency and effectiveness. We also draw conclusions between academic versus commercial annotators.

Search results clustering (SRC) is a challenging algorithmic problem that requires grouping together the results returned by one or more search engines in topically coherent clusters, and labeling the clusters with meaningful phrases describing the topics of the results included in them.

This work investigates the use of linguistically motivated features to improve search, in particular for ranking answers to non-factoid questions. We show that it is possible to exploit existing large collections of question–answer pairs (from online social Question Answering sites) to extract such features and train ranking models which combine them effectively. We investigate a wide range of feature types, some exploiting natural language processing such as coarse word sense disambiguation, named-entity identification, syntactic parsing, and semantic role labeling. Our experiments demonstrate that linguistic features, in combination, yield considerable improvements in accuracy. Depending on the system settings we measure relative improvements of 14% to 21% in Mean Reciprocal Rank and Precision@1, providing one of the most compelling evidence to date that complex linguistic features such as word senses and semantic roles can have a significant impact on large-scale information retrieval tasks.

We present a novel approach to query reformulation which combines syntactic and semantic information by means of generalized Levenshtein distance algorithms where the substitution operation costs are based on probabilistic term rewrite functions. We investigate unsupervised, compact and efficient models, and provide empirical evidence of their effectiveness. We further explore a generative model of query reformulation and supervised combination methods providing improved performance at variable computational costs. Among other desirable properties, our similarity measures incorporate information-theoretic interpretations of taxonomic relations such as specification and generalization.

Web textual advertising can be interpreted as a search problem over the corpus of ads available for display in a particular context. In contrast to conventional information retrieval systems, which always return results if the corpus contains any documents lexically related to the query, in Web advertising it is acceptable, and occasionally even desirable, not to show any results. When no ads are relevant to the user's interests, then showing irrelevant ads should be avoided since they annoy the user and produce no economic benefit. In this paper we pose a decision problem to swing, that is, whether or not to show any of the ads for the incoming request. We propose two methods for addressing this problem, a simple thresholding approach and a machine learning approach, which collectively analyzes the set of candidate ads augmented with external knowledge. Our experimental evaluation, based on over 28,000 editorial judgments, shows that we are able to predict, with high accuracy, when to swing for both content match and sponsored search advertising.

Online photo sharing systems, such as Flickr and Picasa, provide a valuable source of human-annotated photos. Textual annotations are used not only to describe the visual content of an image, but also subjective, spatial, temporal and social dimensions, complicating the task of keyword-based search. In this paper we investigate a method that exploits visual annotations, e.g. notes in Flickr, to enhance keyword-based systems retrieval performance. For this purpose we adopt the bag-of-visual-words approach for content-based image retrieval as our baseline. We then apply rank aggregation of the top 25 results obtained with a set of visual annotations that match the keyword-based query. The results on retrieval experiments show significant improvements in retrieval performance when comparing the aggregated approach with our baseline, which also slightly outperforms text-only search. When using a textual filter on the search space in combination with the aggregated approach an additional boost in retrieval performance is observed, which underlines the need for large scale content-based image retrieval techniques to complement the text-based search.

Sponsored search is one of the enabling technologies for today's Web search engines. It corresponds to matching and showing ads related to the user query on the search engine results page. Users are likely to click on topically related ads and the advertisers pay only when a user clicks on their ad. Hence, it is important to be able to predict if an ad is likely to be clicked, and maximize the number of clicks. We investigate the sponsored search problem from a machine learning perspective with respect to three main sub-problems: how to use click data for training and evaluation, which learning framework is more suitable for the task, and which features are useful for existing models. We perform a large scale evaluation based on data from a commercial Web search engine. Results show that it is possible to learn and evaluate directly and exclusively on click data encoding pairwise preferences following simple and conservative assumptions. We find that online multilayer perceptron learning, based on a small set of features representing content similarity of different kinds, significantly outperforms an information retrieval baseline and other learning models, providing a suitable framework for the sponsored search task.

We discuss the problem of ranking very many entities of different types. In particular we deal with a heterogeneous set of types, some being very generic and some very specific. We discuss two approaches for this problem: i) exploiting the entity containment graph and ii) using a Web search engine to compute entity relevance. We evaluate these approaches on the real task of ranking Wikipedia entities typed with a state-of-the-art named-entity tagger. Results show that both approaches can greatly increase the performance of methods based only on passage retrieval.

No abstract available.