The media has enormous influence on its consumers. With respect to language change, communication studies have shown that television can also be a factor in language change (Stuart-Smith et al., 2013). For example, studies on the language of the Israeli radio broadcasting have noted the improper use and decreasing validity of the Hebrew language, as it is apparent in the last few decades (Marco, 2008). Change are also reflected in the sound patterns of the language. The Hebrew case is of interest, since the first generation of Hebrew speakers (beginning of the 20th century; (circa 1914) could not adopt a concurrent form of naturally spoken Hebrew, as none existed at the time (Amir, Cohen, Reshef, & Gonen, 2019). Hebrew as a native language (i.e., "Mother Tongue") is spoken in Israel for less than a century and from that moment on, Hebrew develops and changes like any other living language (Gonen, 2019).
Broadcast media is almost the only resource of information we have on the early days of spoken Hebrew, and it can show how the language has been evolved, undergoing many technological and cultural changes. Such examination can contribute to sociolinguistic research on the evolvement of the general Israeli accent from two main diasporas (Ashkenazi and Sephardic); and further, to ask if this change can shed light on the hegemonic group that controlled the broadcast media in its early days, compared to present days. I.e., to ask if the change is because of a social transformation. 
Language style can be detected by phonetic and prosodic traits (Stuart-Smith et al., 2013). By using acoustic analysis and audio signal processing methods, we want to explore changes of language style in the Israeli broadcast television. As a first step, we wish to investigate the possibility of carbon-dating (i.e., estimate the accurate period) an audio chunk, based on its acoustic signal. We therefore consider the following question: given speech segments of broadcasting, how well can we automatically classify them according to their periodical style? If native listeners of a language can perceive old speaking styles from present-day style – why not teach the machine to identify it?
In a previous study (Carmi, Lerner, & Silber-Varod, 2019), we succeeded in obtaining good classification between "present-days" and "early-days" of varied sound types. In the current study, we focus only on speaking excerpts of two TV broadcasting periods: 1968-1982 and 2010-present. The first period reflects the emergence of the Israeli Broadcasting Authority (IBA), when only one television channel was available and the IBA was under the enforcement of a dominant prescriptive grammarian – Aba Bendavid; the second period reflects present-day speech in multi-channels era. The dataset is composed of 300 news programs, 150 per each period class, taken from the Kan archive collection (former IBA). The procedure includes Voice Activity Detection (VAD) and extraction of a set of prosodic-based features that were found to be related to speaking style and genre (Belião, 2014; Carmi, Cohen, Avigal1, & Lerner, 2019; Kolár, Liu, & Shriberg, 2009). Results show that machine learning XGBOOST model obtained accuracy of 0.77, indicating a fair prediction of the period class. Several factors might explain these results, apart from style change: Age of the main speakers (young versus old), speaker identification – a set of several unique speakers appear in each video, and the recordings' quality.
Driven by the motivation to trace language style changes, we presented a media archeology (Elsaesser, 2016) study, which links fields of enquiry for mutual verification that would not ordinarily be related by traditional means. Thus, the "iconic" nature of speech and language can become an asset in making sense of our rich linguistic, historic, and metaphorically archaeological past. We demonstrated how digital analysis can reveal more about the sonic dimensions of the language, and about the place of sound in historical understanding more broadly (Kramer, 2018). In future study, we intend to adopt techniques of Distant Listening (Clement et al., 2013), which involve visualization of the audio signal and to move between the optic and the aural through new circuits of computational exploration. As Kramer (2018) argues, this can bring out concealed historical information and to generate more compelling historical interpretations.
As a by-product, this study can offer an automatic enrichment of video databases, beyond the traditional manual meta-data. Audio information retrieval is an emerging field for accurate search results (Pradhan & Sharma, 2018), and innovative tools based on artificial intelligence systems are developed nowadays to tackle the massive amount of audio and video excerpts that need to be indexed (Scagliola, et al., 2019). We suggest that stylistic annotation of broadcast media may be used to automatically index massive amount of cultural heritage videos with no (accurate) metadata concerning their periodical information.

Carbon-dating broadcasting styles: Towards automatic detection of spoken Hebrew change

The international conference #DHJewish - Jewish Studies in the Digital Age will bring together more than 60 scholars and heritage practitioners to discuss how the digital turn affects the field of Jewish Studies. The conference is organised by the [Luxembourg Centre for Contemporary and Digital History (C²DH).](https://www.c2dh.uni.lu/) 

[For more information, and a PDF download of the program, please see also here.](https://www.c2dh.uni.lu/events/dhjewish-jewish-studies-digital-age) 



In order to meet other participants and for informal chats in the breaks, join us in our [wonder.me room](https://www.wonder.me/r?id=73a12756-8d93-4871-9374-143253bfbf59).