Zing computes document scores using three components

  • Release version: Xanadu
  • Updated August 1, 2024
  • 2 minutes to read
  • Summarize
    Summarized using AI
    This content was generated using new OpenAI-powered functionality. Results are provided on an as is basis and are not guaranteed to be accurate or complete.

    Summary of Zing computes document scores using three components

    The Zing search engine calculates document relevance scores based on three key components: frequency, sequence, and weight of search terms within a document. These components work together to rank search results effectively, helping users find the most relevant documents for their queries.

    Show full answer Show less

    Key Features

    • Frequency: Zing counts how often each search term appears in a document, awarding one point per occurrence. This raw frequency can be enhanced by enabling term frequency–inverse document frequency (TF-IDF), which boosts the score of terms that are frequent in a document but rare across the entire document set, improving relevance.
    • Sequence: Zing awards exponentially higher points when search terms appear in the same order as the query. Points are calculated as 10 to the power of the number of sequential terms found (e.g., 10² for two-term sequences, 10³ for three-term sequences). This scoring is multiplied by the field’s weight.
    • Weight (tsweight attribute): Each field in a document has a scoring weight that amplifies the frequency and sequence points. Default weights are higher for important fields such as Knowledge record numbers (50), Knowledge short descriptions (10), and task numbers (50), while all other fields default to 1. The maximum weight is 255.

    Practical Application for ServiceNow Customers

    Understanding how Zing scores documents allows you to optimize search relevance in your ServiceNow instance. You can:

    • Enable TF-IDF to prioritize less common but highly relevant terms, improving search accuracy for your users.
    • Adjust field weights (tsweight) to emphasize critical fields such as record numbers and descriptions, ensuring that matches in these fields contribute more significantly to the score.
    • Leverage sequence scoring to boost results where search terms appear in the order users expect, enhancing user satisfaction with search results.

    These scoring mechanisms combined help deliver more relevant search results, improving findability and efficiency within your ServiceNow environment.

    The Zing search engine computes document scores based on the frequency, sequence, and weight of search terms in the document.

    Document scores

    The components of a document score for a search query are:
    • Frequency: how often the search terms appear in the document.
    • Sequence: how often the search terms appear in the same order as the search query.
    • Weight: how heavily weighted the source field is in which the search terms appear.
    Figure 1. Sample document score computation
    Graphic showing frequency and sequence scoring for sample search query and document.

    Frequency points

    Zing awards one point whenever a search term appears anywhere in the document. For example, when searching for distributed database server, a document that contains distributed three times, database five times, and server 17 times would have 25 frequency points.

    To increase search result scores of search terms that appear more frequently in a document, but less frequently in a document set, you can Score search terms by inverse document frequency (IDF). When TF-IDF is enabled, search term scores are calculated by multiplying the term frequency score by the inverse document frequency score. Because enabling TF-IDF increases the weight of less common search terms, search results for that table are more likely to be relevant. For example, when searching for distributed database server, the term distributed might receive a higher score than server if it appears frequently in one document but less frequently in the document set as a whole.

    Zing applies a multiplier to frequency points based on the value of the ts_weight attribute for the field in which the search term appears. A field with a text search scoring weight of 30 (ts_weight=30) would add 30 points for each inclusion of a search term.

    Sequence points

    Zing awards a document more points when it contains the search terms in the same order in which they were typed. The more search terms in sequence there are, the exponentially higher the score becomes. Zing awards sequence points as 10^x, where x is the number of search terms that appear in sequence.

    In the distributed database server search example, Zing awards a document 100 (10^2) sequence points for each time it includes the two-term string database server. Likewise, Zing awards a document 1000 (10^3) sequence points each time it includes the three-term string distributed database server.

    Zing applies a multiplier to sequence points based on the value of the ts_weight attribute for the field in which the sequence appears. The sequence points use the calculation (10^x * field ts_weight attribute).

    Field scoring weights

    The system elevates the default scoring weight of Knowledge record numbers, Knowledge short descriptions and metadata, task record numbers, and task short descriptions. Default ts_weight attributes for these fields are as follows:
    • kb_knowledge.number = 50
    • kb_knowledge.short_description = 10
    • kb_knowledge.meta = 10
    • task.number = 50
    • task.short_description = 10

    All other fields have a default ts_weight attribute of 1. The maximum possible weight value is 255.