Tag strictness and match rates

When memoQ looks up a segment in a TM or a LiveDocs corpus, tag permissiveness settings may allow exact matches even when the inline tags are not exactly the same in the source segment and the text that was found in the TM or LiveDocs corpus. You can change the settings of tag strictness in TM settings or LiveDocs settings, using the Edit TM Settings dialog or the Edit LiveDocs settings dialog.

See also: To get an overview of match rates in general, read the topic about Match rates from translation memories and LiveDocs corpora.

  1. If tag strictness is set to strict:

    The score is not adjusted, only segments with identical tag properties get a 100% score. Tag properties are: tag position, tag type (one of opening/closing/empty/memoQ), tag name, and tag attribute names and values.

  2. If tag strictness is set to medium:

    memoQ will report an exact (100%) match if the tags are the same type and they occur in the same position as in the source text. They can have different attributes; that will not result in a lower match rate.

    3. Examples:

    1. Take this segment from the source text:
      sample0-source

      There's a match for it in the TM or in the LiveDocs corpus:
      sample0-match

      memoQ will report an exact (100%) match because the two segments differ only in the attribute value of the <a> tag.

    2. On the other hand, look at this segment:
      sample1-source

      memoQ returns this match:

      sample1-match

      This will not be an exact match because the order of the <a> and <strong> tags is different in the two segments.

    3. memoQ will still return an exact (100%) match if the segment in the TM or the LiveDocs corpus contains extra tags (tags that are not in the new source segment). In other words, the TM or the LiveDocs corpus must have a tag for each tag in the source segment, but it is not a problem if the source segment does not have a matching tag for each tag in the TM or LiveDocs corpus. For example, this will return an exact match:
      Source segment:

      sample2-source

      Match for the TM or LiveDocs corpus:

      sample1-source

  3. If tag strictness is set to permissive:

    memoQ will return an exact match if it finds a hit that has a tag or a tag sequence everywhere where there is a tag or tag sequence in the source segment. The types or the attributes of the tags do not make a difference.

    Again, memoQ will still return an exact (100%) match if the TM or LiveDocs match has tags that the current source segment does not have. memoQ will silently ignore the extra tags in the TM or LiveDocs match.

    Example:

    This is the source segment:

    sample3-source

    And this is the match from the TM or the LiveDocs corpus:

    sample3-match

    If tag strictness is set to permissive, this will be an exact (100%) match.

Adjustment of target tags when a hit from the TM is inserted

When a TM hit is inserted into the translation document, memoQ tries to adjust the tags in the TM target segment to make them as close to the tags of the looked up source as possible.

If the lookup segment has exactly the same tag information as the TM segment (in the strict sense), then

  • no adjustment is needed. The TM target segment is copied into the translation document as is.

    If the lookup segment's tag information is different from that of in the TM (the score is less than 100% or it is 100%, but the tag strictness is set to medium or permissive),

  • memoQ tries to find the corresponding tag pairs in the lookup and the TM source segment.
  • memoQ tries to find the pair of each TM source segment tag in the TM hit's target segment.
  • memoQ replaces the tag in the TM hit's target with the tag in the lookup source.

Example:

Lookup of segment:

<A>text<B/></A>

TM entry found:

source:  <C>text</C>

target:  <C>texto</C><D></D>

pairing of tags between lookup and TM source:

<A>  →  <C>

<B/>  →  nothing

</A>  →  </C>

pairing of tags between TM source and TM target:

<C>  →  <C>

</C>  →  </C>

nothing  →  <D>

nothing →  </D>

adjustment of the target, when inserted:

<C>  →  <A>

</C>  →  </A>

<D></D>  →  remains the same

so the adjusted target segment would be:

<A>texto</A><D></D>

In most cases, the tags in the TM source and TM target are identical. But if the TM target has some extra tags (<D></D> in the above example), those are copied into the translation document unchanged.