Report:Delphion/Search Syntax/Allowed Operators/Boolean and Proximity Operators

The search algorithm is not case-sensitive, so operators do not have to be input in all capital letters.

Proximity/Positional Operators

The first type of proximity operator that can be employed in Delphion is simply leaving a blank space between words. Delphion performs “phrase searching” by default; in other words, a blank space equals an adjacency operator, where words will be searched in sequence, the first word must appear directly before the second, and automatic stemming is only applied to the last word in the phrase.

Phrase searching is always on in Delphion. To negate it, or “turn it off” as the help file says, users must construct their queries indicating the desired relationships between keywords with Boolean or proximity operators.

Although it offers many field-limited search options via its advanced search form, experienced users will likely prefer Delphion’s CLI-style language to input advanced query logic. Searching this way can support field-specific text queries, and can even support text queries with weighting or proximity requirements.

In order to enter a field-specific query in the Delphion query language, users must become familiar with the field-designating operator, <in>;. Delphion describes the operator as a “proximity” operator, but actually it is only used to designate specific text data fields for search, such as title, abstract, assignee, etc. (the <in> operator does not apply to date fields).

The <in> operator works with either the full, or Delphion-specific abbreviation, of a data field name. For example:

(Velcro <in> title); or, (Velcro <in> ti)

Are both valid queries. A complete list of field abbreviations used by Delphion (taken from the Delphion help file) is included at the end of this section.

To search for two words within a defined proximity to each other, Delphion offers the three operators, <near> <near/n> and <order>. These operators are actually quite different from “near” or “adj” operators in other systems:

<near> will retrieve any document that contain the two terms, and will rank the results based on the proximity of the keywords to one another. The system scores a zero for documents in which the two terms specified occur within greater than 1024 words of one another.

A few things are unusual about this operator: firstly, 1024 words is a very high threshold for a proximity operator to allow; secondly, it’s unusual that the query will still return any results containing the two keywords, whether or not they occur within 1024 words of each other. So, the <near> operator would appear to return the exact same document set as the same query, using the Boolean AND operator instead of <near>. The fundamental difference is that the relevancy ranking method of the two results sets would be different. For example, compare the strings (dog and collar) vs. (dog <near> collar): The first query’s hit list (the AND query) would be ranked as a function of the number of occurrences of the words “dog” and “collar” over the total length of the document; in contrast, the second query’s hit list (the <near> query) would be ranked (partially) as a function of the proximity of the two words to each other.

<near/n> will retrieve any document that contains the two keywords within the designated proximity to one another, where n can be any integer up to 1024.

This operator is much more akin to the “near” or “adj” operators offered by some other search engines, meaning that the results set will only contain those documents which actually contain the keywords specified within the desired proximity. Relevancy rankings will still be based upon how close the two keywords actually are to each other; however, the smaller the allowed proximity range, the less variation in relevancy rankings can be observed. For example, results returned when n was 4 ranged from 100% relevant (the two words were adjacent) to 97% relevant (the two words were 4 words apart).

<order> is an operator that must be used in conjunction with <near> or <near/n>, and requires that the two terms must be found occurring in the order they appear in the query itself.

The <order> operator exists to increase search accuracy when one keyword should always occur before the other. Competing search engines satisfy this need in a number of different ways, from defining two separate proximity operators (for example, MicroPatent PatentWeb’s ADJn searches for keywords in sequence, while NEARn searches for the keywords in either order), to allowing a single letter to be inserted into a particular command to specify when sequence is important (for example, PatBase’s “wn” operator is for keywords in unspecified sequence, whereas “wfn” denotes fixed sequence searching). Delphion’s way of solving the problem by adding a separate operator is just another method; although some might find having to type an extra word in brackets cumbersome, others may find this command easy-to-remember and more intuitive than other systems.