AKS: Focusing your Search

Sometimes it is difficult to know how to structure a search. If a search is very unspecific it retrieves more documents than we can handle. If, on the other hand, a search is too specific little or no documents are retrieved. This is no problem with AKS, since it automatically analyzes the search and proposes keywords to refine and focus the search according to specific needs.

Example
This example helps to illustrate how a search can be gradually improved to eventually obtain more concrete and valuable results.

Fig 1. List of terms significantly linked to leptin. There appears to be a strong connection between leptin and obesity. This set of keywords helped us narrow our search.

We initiate a search to retrieve all documents mentioning the gene "Leptin". We perform an analysis of the document set and check for the relevant terms (see Fig. 1, for more information on document analysis see Large document set analysis).

Fig 2. Document list from the "leptin and obesity" document set ranked by their informative relevance. Some of them [documents no. 1, 9, 11 and 12] suggest the relationship. A subset of the documents [documents no. 1,7, 13 and 21] suggests something new: energetic aspects are involved in the relationship "leptin-obesity".

The presence of word-root "obes" led us to believe it is linked to obesity so we perform a new analysis looking for a more restrictive set of documents that contains both "leptin" and "obesity".

In this analysis result we take a look at the document and sentences ranked according to their informative relevance within the document set (Fig. 2). We find that some of the documents support the relationship and that this relationship becomes more evident if we take a look at the sentences (Fig. 3). Moreover, we can see that some of the documents and sentences emphasize a relationship of the leptin with diverse aspects of the energetic metabolism.

Fig 3. Sentence list from the "leptin and obesity" document set ranked by their informative relevance. Seventeen out of the 30 top-ranked sentences supported the relationship.

Looking at the terms extracted for the analysis of a combined search of leptin and obesity (Fig. 4), we can see that word-roots related to energy like "energi" and "thermogenesi" appear.

Focusing on the bigrams we can see consecutive terms such as "energi homeostasis" or "energi balance" which lead us to the same conclusion (Fig. 5).

To focus the study in the metabolism, we select some terms related to obesity and energy and then click on "Update active terms" button. Then we focus on bigrams (consecutive terms) and select those regarding energy metabolism. We click again on"Update active terms" and then we start the new ranking procces by clickin on "Start process". The system will re-calculat the relevance scores for sentences, documents and authors taking into account the selected items (Figs. 4 & 5).

We check again for the most relevant documents and sentences (Fig. 6 & 7) holding the energetic aspects of "leptin & obesity" document set. We apply the filter to focus directly on the sentences mentioning "obesity" (Fig 8).

Fig 8. New list of ranking sentences after applying a filter to retrieve those containing the word "obesity". Sentences no. 3, 4, 5, 6, 17 and 19 clarify the relationship leptin-metabolism-obesity. New ideas could be studied arising from this list, for instance, the relationship with insulin resistance and the onset of diabetes (sentences no. 13, 17, 20, 21, 22 and 29).

This way, we progressively focus the search towards the aspects we consider interesting until we get to a set of documents and sentences that are highly informative for a complex criterion: "relation of leptin and obesity, considering specially the aspects concerning energetic metabolism".