EPTIC

Documentation

The European Parliament Translation and Interpreting Corpus (EPTIC) is developed at the University of Bologna and a few other universities responsible for different language components. At present EPTIC comprises texts in English, Italian and French; however, other language components, including Finnish, Polish and Slovene are currently on the way.

The corpus is currently made available through the NoSketch Engine platform (NoSke; Rychlý 2007).

EPTIC is an intermodal and parallel corpus of a complex structure. Data included in the corpus is derived from the official website of the European Parliament, which provides videos and verbatim reports of the plenary sessions together with the interpretations of the speeches, as well as their translations (the latter only for plenaries which took place until mid-2011). Subcorpora are aligned to each other at sentence level, and transcripts of speeches and interpretations are time-aligned with the corresponding videos.

Each language combination component in EPTIC includes the following subcorpora:

• sources – spoken: orthographic transcripts of the original speeches;
• sources – written: official verbatim reports of the source speeches;
• targets – interpreted: transcripts of the interpretations;
• targets – translated: translations of the verbatim reports.

EPTIC is also a multilingual corpus, which includes several language combinations and translation/interpreting directions. All language combinations feature English as one of the languages involved. The subcorpora completed so far are: English>French, French>English, English>Italian, Italian>English, Polish>English.

Sizes of individual language components

The current sizes of individual language components of EPTIC are visible in the following Tables (count of tokens obtained via the Corpus info function of NoSkE).

	Sources		Targets
	Spoken	Written	Interpretations	Translations
English	24,136	22,782	53,615	58,561
French	27,713	26,674	23,185	25,855
Italian	20,016	19,591	20,352	23,234
Polish	11,011	10,616	TBA	TBA

The English subset of EPTIC has a slightly different structure from the remaining ones. It comprises source texts originally delivered by native and non-native speakers as well as translations and interpretations from various languages. Since little information is provided at the European Parliament website about the translations into English, it is difficult to establish whether they are produced by native English speakers or not. The interpretations into English from various languages are carried out either by native or non-native English speakers.

English
Spoken sources		Written sources		Interpretations			Translations
Native	Non-native	Native	Non-native	From French	From Italian	From Polish	From French	From Italian	From Polish
									into A	into B
10,840	13,296	10,312	12,470	23,670	18,390	11,555	25,046	20,335	13,180	NA

Characteristics of EPTIC texts

EPTIC is comprised mostly of short speeches ranging from 100 to 300 words, but there are also medium and long speeches (the latter exceeding 1000 words). Speeches are delivered at the European Parliament plenary sessions by different speakers, including mostly MEPs, but also commissioners and guests. They are devoted to a range of topics (see Table below). The composition of the individual subcorpora with respect to the topics covered in the debates can be inferred by clicking on the Corpus info link within the NoSke interface: in the Structures and attributes section, simply click on Topic.

Topics covered in EPTIC
Agriculture and Fisheries Economics and Finance Employment Environment Health Justice Politics Procedure and Formalities Society and Culture Science and Technology Transport

Filters (metadata) available in EPTIC

While searching the corpus, it is possible to filter the queries using the contextual information that allow to narrow down the query to, e.g., a speech delivered by a particular speaker, or a speech of particular length. The filtering options are described in the following Table.

As the corpus is aligned at sentence level, it is possible to search all 4 aligned subcorpora in a parallel search. In addition, the corresponding excerpt of a video of the source speech or the interpretation can be displayed.

Filter	Description
text.id	refers to the ID of the text
text.date	date on which the speech was delivered at the EP
text.length	length of the speech in general (short, medium, long)
text.lengthw	exact text length in words
text.duration	duration of the speech (short, medium, long)
text.durations	duration of the speech in seconds
text.speed	refers to the pace of delivery of the speech (low, medium, fast)
text.speedwm	speed of delivery expressed in words per minute
text.delivery	read vs. impromptu
text.topic	the general topic of the speech
text.topicspec	title of the debate
text.type	source-spoken/ source-written/ target-interpreted/ target-translated
text.wordcount	length of the speech in words
speaker.name	name of the speaker
speaker.gender	gender of the speaker
speaker.country	country the speaker represents
speaker.native	the speaker is speaking the native tongue or a foreign language
speaker.politfunc	the speaker’s political function at the EP
speaker.politgroup	the speaker’s political group
st.language	source text language
st.length	source text length
st.lengthw	source text length in words
st.duration	source text duration in general (short, medium, long)
st.durations	source text duration in seconds
st.speed	pace of the original speaker in general (low, medium, fast)
st.speedwm	pace of the original speaker in words per minute
st.delivery	mode of delivery of the source speech (read vs. impromptu)
interpreter.gender	gender of the interpreter
interpreter.native	the interpretation is delivered into the native tongue or the foreign language