Document Translation

In order to translate a document with the Lilt API, you need to do the following steps:

1. Upload a file to a project (currently, for API upload we only support xliff, html, xlsx, csv)

   # Python; Note: This is just a stub
   def upload_document(filename, projid):
    payload = {"key": lilt_api_key}
    jsonData = {"name": fileName, "project_id": projid}
    headers = { "LILT-API": json.dumps(jsonData), "Content-Type": "application/octet-stream" }
    with open(fileName, 'r') as fp:
        rawData = fp.read()
    res = requests.post(lilt_api_url + "/documents/files", params=payload, data=rawData, headers=headers, verify=False)
    return res.json()["id"]
   

2. Retrieve segment source and ids

   # Python; Note: This is just a stub
   def get_document(docid):
    segments = []
    payload = {"key": lilt_api_key, "id": docid}
    res = requests.get(lilt_api_url + "/documents", params=payload, verify=False)
    for seg in res.json()["segments"]:
        segments.append(seg)
    return segments
   

3. Translate each source segment

   # Python; Note: This is just a stub
   def translate(source, memid):
    payload = {"key": lilt_api_key, "memory_id": memid, "source": source, "tm_matches": "false"}
    res = requests.get(lilt_api_url + "/translate", params=payload, verify=False)
    return res.json()[0]
   

4. Update each segment with the translation

   # Python; Note: This is just a stub
   def update_segment(segid, target):
    payload = {"key": lilt_api_key}
    jsondata = {"id": segid, "target": target}
    res = requests.put(lilt_api_url + "/segments", params=payload, data=jsondata, verify=False)
   

5. Download the file

   # Python; Note: This is just a stub
   def download_document(fileName, docid):
    payload = {"key": lilt_api_key, "id": docid}
    res = requests.get(lilt_api_url + "/documents/files", params=payload, verify=False)
    with open(fileName, 'wb') as fp:
        fp.write(res.content)
   

To adapt your machine translation engine and update the TM in your Memory you should also add the new segments with their final correct translation (after verification/processing by a translator or bilingual) to the Memory.

Segmentation

Lilt performs sentence segmentation on source segments. To bypass the segmentation, add a <seg-source> element which indicates a segmented source, and corresponding <mrk> markers inside the segment to specify the segment boundaries. Note that you have to put both, <source> and <seg-source>, even if ultimately the latter overrides the former. For example:

<trans-unit>
    <source>Segment this source. Any content here will be imported as multiple segments in Lilt. Try it out!</source>
    <seg-source><mrk mtype="seg">Do not segment this source. Any content here will be imported as one segment in Lilt. Try it out!</mrk></seg-source>
    <target />
</trans-unit>

To learn more, see the full API reference.