New Re-Ranking Approach in Merging Search Results

Abstract

When merging query results from various information sources or from different search engines, popular methods based on available documents scores or on order ranks in returned lists, its can ensure fast response, but results are often inconsistent. Another approach is downloading contents of top documents for re-indexing and re-ranking to create final ranked result list. This method guarantees better quality but is resource-consuming. In this paper, we compare two methods of merging search results: a) applying formulas to re-evaluate document based on different combinations of returned order ranks, documents titles and snippets; b) Top-Down Re-ranking algorithm (TDR) gradually downloads, calculates scores and adds top documents from each source into the final list. We propose also a new way to re-rank search results based on genetic programming and re-ranking learning. Experimental result shows that the proposed method is better than traditional methods in terms of both quality and time. Povzetek: V prispevkih sta primerjana dva pristopa pri združevanju zadetkov iskanja: z enačbo in z algoritmom TDR, nato pa je primerjana še izvirna metoda.

Machine learning methods for results merging in patent retrieval

“…In terms of download methods, Hung (2019) proposed a technique in which the best documents are downloaded to re-rank and create the final merged list. He used ML and genetic programming to re-rank the final merged results.…”

“…are used to calculate the relevance of the documents returned from the heterogenous remote collections (Callan et al , 1995; Si and Callan, 2003; Shokouhi and Zobel, 2009). The download methods, download all the documents returned and re-calculate their relevance to the query locally (Craswell et al , 1999; Hung, 2019). Finally, hybrid methods are a combination of estimation and download methods (Paltoglou et al , 2008).…”

Using Ant Colony Optimization for Results Merging in Federated Search

“…The collection selection [5,6] uses the sampled information to select a few collections that may contain relevant documents against a searc query. Finally, the results merging [7,8] combines the results from the component collections into an ordered list.…”

“…Learning to rank (L2R) techniques have been considered an alternative to traditional document-ranking approaches in recent years. Several research works [7,[9][10][11][12][13] have used L2R for various retrieval tasks. In these techniques, a set of queries, each associated with the ranked list of documents and their relevance judgments, is exploited as training data [14].…”

“…These techniques have been reported [15] to provide better solutions than the algorithms de ned by experts. A recent study [7] used genetic programming to solve the result merging problem in which returned documents by the collections are downloaded, some likely relevance features are extracted from the documents at each iterative process, and then the documents are re-ranked based on their relevance scores. However, the drawbacks of this approach are increases in bandwidth usage, computational resources, and latency time.…”