Product DocsMenu

Customizing Search Results Ranking

CES uses 18 ranking factors divided into 6 types to calculate the relevancy score of documents. CES natively uses pre-tuned ranking weights that are likely to be satisfying most of the time. Nevertheless, fine-tuning these factors can sometimes increase accuracy, especially for specific situations.

Example: An organization may find not important that a recent document modification date contribute to a higher rank.

As a Coveo administrator, you can fine-tune the ranking of search results by adjusting the weight of each ranking factor and specify ignored fields, fields whose values are not taken into consideration when ranking results (see To customize the weight of ranking factors). By default, each of these factors is assigned a weight of 5 on a scale of 1 to 9, where 9 corresponds to the highest ranking weight.

One of the CES ranking factor is the collaborative rating feature that allows users to voice their personal appreciation of documents (see Rating a Search Result). Collaborative rating is enabled by default. You can disable it or configure who can use it and on to how many Top Result documents it applies (see To customize collaborative rating parameters).

To customize the weight of ranking factors

  1. On the Coveo server, access the Administration Tool (see Opening the Administration Tool).

  2. Select the Index > Ranking.
  3. In the panel on the left, select Ranking Weights.

  4. In the Ranking Weights page, for each factor:

    • Select a higher value to increase the importance of a factor.
    • Select a lower value to decrease the importance of a factor.
    • Select Ignore to exclude a ranking factor from the ranking process.

    Note: Even if all factors have the same weight (ex.: 5), their relative importance is not necessarily identical as CES uses a built-in weight for each factor. For example, Term in title has more effect on the final ranking than Document modified recently.
    Each ranking factor value is relative to the other factor weights. The resulting ranking is the same whether all multipliers are set to 4 or if they are all set to 2. You can give a higher weight to a factor by setting it to 7 and all others to, for example, 2. The relative weight of 7 against 2 is much higher than 7 against 4.

    Refer to the table that follows for details on each ranking factors.

    Type Factor Description
    Content Term frequency

    Higher relevancy score to documents in which query terms are frequently repeated. The importance increases proportionally to the number of times a word appears in the document, but is offset by the frequency of the word in the corpus.

    Example: A document in which a term is repeated 10 times receives a higher ranking than one in which it is repeated 5 times.

    Term proximity

    Higher relevancy score to documents in which query terms are close to each other. This factor applies only when the query contains more than one term.

    Example: A document in which the terms are only one word apart receives a higher ranking than one in which they are ten words apart.

    Term in title

    Higher relevancy score to documents containing query terms in their titles. Moreover, relevancy is dependent on the proportion of title terms matching the query. 

    Example: If the query term is Coveo, a document with the title Coveo Enterprise Search (one out of three words matches the query) receives a higher ranking than one with the title Enterprise Search, or Coveo Enterprise Search Provides Businesses with the Most Collaborative and Agile Search Solution Available (one out of 14 words matches the query).

      Term in concepts

    Higher relevancy score to documents containing query terms in their lists of concepts. Concepts are extracted by the CES linguistic algorithm which relies on term frequency and proximity.

    Example: If the query is Coveo, a document for which the concept Coveo has been extracted receives a higher ranking than one for which this concept has not been extracted (even if the word Coveo is present in its content).

    Term in summary

    Higher relevancy score to documents containing query terms in their summaries. Summaries are extracted by the CES linguistic algorithm which relies on term frequency and proximity (see What Are Summaries?).

    Example: If the query term is Coveo, a document for which the key topic Coveo has been extracted in its summary receives a higher ranking than one for which it has not been extracted (even if the word Coveo is present in its content).

    Term matches original word (accents, casing and stemming)

    Higher relevancy score to documents in which query terms correspond exactly to the words found in the text. This factor is fully effective only if original words, accents and casing are indexed (see How to Modify the Index Settings).

    Example: If the query term is COVEO, a document containing the word COVEO receives a higher ranking than one containing Coveo.

    Document Documents modified recently

    Higher relevancy score to documents whose modification date is recent.

    Example: A document modified today receives a higher ranking than one modified a month ago.

    Document quality evaluation

    Higher relevancy score to documents which are close to the root of a file system.

    Example: A document located in C:\CES6\ receives a higher ranking than one located in C:\CES6\Index\Default.

    Document in user language

    Higher relevancy score to documents in the same language as the search interface.

    Example: If the search interface is in English, a document written in English receives a higher ranking than one written in French.

    Location Source rating

    Higher relevancy score to documents contained in a source rated Highest to Above Normal as you can individually set the rating of each source (see Adding a Source).

    Terms in address

    Higher relevancy score to documents whose paths contain query terms.

    Example: If the query term is Coveo, a document whose address is http://www.coveo.com/en/Support/Default.aspx receives a higher ranking than one whose address is http://support.microsoft.com/.

    Term in folder name

    Higher relevancy score to documents whose parent folders name contain query terms.

    Example: If the query term is Coveo, a document located in the Coveo Help folder receives a higher ranking than one located in Help.

    Term in home page address

    Higher relevancy score to home pages whose addresses contain a query term.

    Example: If the query term is Coveo, a document whose home page is http://www.coveo.com receives a higher ranking than one whose home page is http://support.microsoft.com/.

    Formatting Term in large fonts

    Higher relevancy score to documents containing query terms in large fonts or bold.

    Example: If the query term is Coveo, a document containing the word Coveo receives a higher ranking than one containing the word Coveo.

    Term is underlined

    Higher relevancy score to documents containing underlined query terms.

    Example: If the query term is Coveo, a document containing the word Coveo receives a higher ranking than one containing the word Coveo.

    Collaborative Collaborative rating weight

    Higher relevancy score to documents whose collaborative rating score is high (see What Is Collaborative Rating?). Collaborative rating must be enabled for this factor to apply (see To customize collaborative rating parameters).

    Custom Custom ranking weight

    Determines the importance that must be given to documents whose custom ranking weight has been set. Custom ranking factors are scripted directly in open converters (see What Are Converters? and Adding an Additional Connector).

  5. In the Ignored Fields text box that already contains built-in fields that should be excluded, add the custom fields that you wish to exclude from the ranking results (see What Is the Difference between Built-In and Custom Fields?).

    Example: If the query expression is Coveo AND @syslanguage=English, occurrences of the term English do not affect ranking because syslanguage is an ignored field. However, if syslanguage is removed from the Ignored Fields list, ranking is adversely affected because documents containing occurrences of the word English receive a higher ranking. Whereas, the purpose of syslanguage is to find documents written in English, not to display all documents with the word English in them. If custom fields can have a similar effect on ranking, add them to the Ignored Fields list.

  6. Click Reset below the label of a factor category, to return the weight of ranking factors for that category to their default value of 5.
  7. Click Apply Changes to save your modifications.

To customize collaborative rating parameters

Calculating collaborative ratings may require considerable CPU resources and can slow down CES. If this problem occurs, consider reducing the number of top documents to which it applies (ex.: 50), reducing the number of groups or simply disabling collaborative rating.

  1. On the Coveo server, access the Administration Tool (see Opening the Administration Tool).

  2. Select the Index > Ranking.
  3. In the panel on the left, select Collaborative Rating.
  4. In the Collaborative Rating page:
    1. To disable the collaborative rating feature, on the Status line, click [Disable]. You will be prompt to confirm this actions as it will delete all the personal appreciations entered so far by users. To re-enable the feature, click [Enable] that appears at the end of the Status line.

    2. On the Option line, you can change the number of top ranking documents on which collaborative rating is available. The default value is 100, meaning that users can set a personal appreciation only on the first 100 top ranking search results documents.

    3. In the Valid Groups list, you can control who can use collaborative rating by specifying groups. By default everyone can use collaborative rating.
      1. To remove a group, select it in the Valid Groups list, and then click Remove.
      2. To add a group, type a valid group name in the text box, and then click Add.
    4. Click Apply Changes to save your changes.