• Home
• Grams
  • N-Grams Simple
  • N-Grams Advanced
  • Phrase-Frames
  • Char-Grams
  • PoS-Grams
  • Simple / RegEx Search Tabbed Output
  • "Drill-down" RegEx Search
• Documentation
  • Getting Started Tutorial
  • FAQ
  • Conventions
  • Part of Speech Tags
  • Fused Forms & Multiword Units
  • Character Entity Remapping
• Other
  • Recent Changes
  • Links to Related Resources
  • Type & Token Length Distribution

Drill-down queries find longer phrases incorporating the query string. The results indicate the number and percentage of occurrences of each phrase accounted for by individual longer phrases. These "expansions" also help detect the boundaries of recurrent phrases.

By default, drill-down queries…

• match all strings containing the string you enter, so trip matches trip, trips, tripped, triple, striped…
• return expansions both right and left of the query string

To restrict the scope or direction matched, insert start / end of word / phrase codes:

• match exact word –  [[:<:]]trip[[:>:]]   only matches phrases which contain the word trip (not tripped, striped etc.), and  [[:<:]]trips?[[:>:]]   matches either trip or trips.
• expansion to right –  ^by[[:>:]]   matches phrases starting in the word by
• expansion to left –  [[:<:]]floor$   matches phrases ending in the word floor

Drill-down results sum all counts of a given string, ignoring PoS tags.  Consequently frequencies may be greater than those reported by other query types when the same sequence of words has been tagged in different ways.

Results are displayed as nested boxes containing all expansions which meet the user's frequency threshold. From the lowest value of n to the highest, each expansion of matching phrases by one word is nested one level deeper.

Take for example the query  [[:<:]]trip[[:>:]]  for n values 2-6 (partial results below, click here to view a larger sample).  The most frequent two-word match is trip to.  This "top-level node" (lowest value of n) appears in the outermost box (solid lines).   Three-word expansions of trip to such as a trip to are nested one level lower in dotted-line boxes within the box of their "parent node" trip to.  In turn a trip to  has its own four-word "child nodes" such as a trip to the, on a trip to nested in a dotted-line box one more level down.  Sibling nodes of a trip to like trip to the and the trip to are nested at the same level under the top-level box.  

The boxes are like modified Venn diagrams using rectangles instead of circles.  Each box contains a subset of the occurrences of the parent phrase:  the set of phrases containing trip to is a subset of all occurrences of trip, and trip to the is a subset of all occurrences of trip to. A phrase's level of nesting reflects the number of supersets above it:  trip ⊃ trip to ⊃ trip to the ⊃ trip to the moon  (⊃ 'superset of').  

Each parent may have many children, i.e. each set may have many subsets.  For orientation in a long list, hover the mouse over a phrase to display the parent phrase and its frequency. To help identify significant recurrent phrase,s "children" are color-coded if they account for more than a given percentage (specified above under "Options") of all occurrences of either the top-level "ancestor" node or their parent node (shades of red and green respectively). 

Child phrases are displayed in decreasing order of frequency.  Five columns of numbers follow each phrase:

1. frequency of the phrase 
2. % of parent node's (= direct superset) frequency it accounts for
3. cumulative % of column 2 for left-expanding phrases (> if not applicable)
4. cumulative % of column 2 for right-expanding phrases (< if not applicable)
5. % of top-level node's frequency it accounts for

Fetching explanation of Regular Expressions